ACTA THE ROLE OF POLYCYSTIC OVARY SYNDROME (PCOS) AND OVERWEIGHT/OBESITY IN WOMEN S METABOLIC AND CARDIOVASCULAR RISK FACTORS AND RELATED MORBIDITIES

Transkriptio

1 OULU 2019 D 1513 ACTA Meri-Maija Ollila UNIVERSITATIS OULUENSIS D MEDICA THE ROLE OF POLYCYSTIC OVARY SYNDROME (PCOS) AND OVERWEIGHT/OBESITY IN WOMEN S METABOLIC AND CARDIOVASCULAR RISK FACTORS AND RELATED MORBIDITIES UNIVERSITY OF OULU GRADUATE SCHOOL; UNIVERSITY OF OULU, FACULTY OF MEDICINE; OULU UNIVERSITY HOSPITAL

2

3 ACTA UNIVERSITATIS OULUENSIS D Medica 1513 MERI-MAIJA OLLILA THE ROLE OF POLYCYSTIC OVARY SYNDROME (PCOS) AND OVERWEIGHT/ OBESITY IN WOMEN S METABOLIC AND CARDIOVASCULAR RISK FACTORS AND RELATED MORBIDITIES Academic dissertation to be presented with the assent of the Doctoral Training Committee of Health and Biosciences of the University of Oulu for public defence in Auditorium 4 of Oulu University Hospital, on 7 June 2019, at 12 noon UNIVERSITY OF OULU, OULU 2019

4 Copyright 2019 Acta Univ. Oul. D 1513, 2019 Supervised by Docent Laure Morin-Papunen Docent Terhi Piltonen Reviewed by Professor Risto Kaaja Docent Oskari Heikinheimo Opponent Professor Joop S. E. Laven ISBN (Paperback) ISBN (PDF) ISSN (Printed) ISSN (Online) Cover Design Raimo Ahonen JUVENES PRINT TAMPERE 2019

5 Ollila, Meri-Maija, The role of polycystic ovary syndrome (PCOS) and overweight/ obesity in women s metabolic and cardiovascular risk factors and related morbidities. University of Oulu Graduate School; University of Oulu, Faculty of Medicine; Oulu University Hospital Acta Univ. Oul. D 1513, 2019 University of Oulu, P.O. Box 8000, FI University of Oulu, Finland Abstract Polycystic ovary syndrome (PCOS) is the most common endocrine disorder affecting reproductive aged women, with reproductive, metabolic and cardiovascular implications across the life span. The typical features of PCOS include irregular menstruation, androgen excess and polycystic ovaries in ultrasonography. The majority of women with PCOS are overweight or obese, and, at least partly, obesity-driven metabolic abnormalities often coexist with PCOS. Despite intensive research, it has remained unclear whether PCOS per se is a risk factor of metabolic abnormalities, and cardiovascular disease and events. The main aim of the current work was to investigate whether PCOS is an independent risk factor of metabolic abnormalities and cardiovascular diseases. The study population consisted of the prospective population-based Northern Finland Birth Cohort 1966, and we used data collected at ages 14, 31 and 46. The definition of PCOS was based on self-reported PCOS symptoms at age 31 and/or PCOS diagnosis by age 46. The results revealed that weight gain in early life was a risk factor for the development of PCOS. As for metabolic outcomes, at age 46, normal-weight women with PCOS did not display increased odds of abnormal glucose metabolism. However, weight gain during early adulthood was significantly associated with abnormal glucose metabolism in women with PCOS by age 46. Interestingly, PCOS per se was already associated with elevated blood pressure at age 31 and hypertension at age 46, independently of obesity. Women with PCOS also displayed reduced cardiac vagal activity, which was associated with metabolic abnormalities and hypertension. Furthermore, even though no major anatomical or functional impairments were observed in echocardiography, women with PCOS displayed a significantly greater prevalence of myocardial infarction and a two-fold higher prevalence of cardiovascular events than controls. In conclusion, our findings indicate that even though PCOS is an independent risk factor of metabolic derangements, related obesity is a major metabolic risk factor in these women. The role of PCOS in cardiovascular events per se remains controversial and requires follow-up of this cohort. Given all this, maintaining normal weight and preventing weight gain, especially during early adulthood, should be the main priority in the prevention of adverse metabolic changes in women with PCOS. Keywords: autonomic nervous system, body mass index, cardiovascular diseases, hyperandrogenism, hypertension, metabolism, polycystic ovary syndrome, prediabetes, type 2 diabetes

6

7 Ollila, Meri-Maija, Munasarjojen monirakkulaoireyhtymän ja ylipainon merkitys naisten metabolisiin ja kardiovaskulaarisiin riskitekijöihin ja tauteihin. Oulun yliopiston tutkijakoulu; Oulun yliopisto, Lääketieteellinen tiedekunta; Oulun yliopistollinen sairaala Acta Univ. Oul. D 1513, 2019 Oulun yliopisto, PL 8000, Oulun yliopisto Tiivistelmä Munasarjojen monirakkulaoireyhtymä (polycystic ovary syndrome, PCOS) on lisääntymisikäisten naisten yleisin hormonaalinen häiriö aiheuttaen runsaasti sairastavuutta ja terveydenhuollon kustannuksia. PCOS:n diagnostisiin kriteereihin kuuluvat epäsäännöllinen kuukautiskierto, lisääntynyt miessukupuoli-hormonivaikutus sekä monirakkulaiset munasarjat. Merkittävä osa oireyhtymää sairastavista naisista on ylipainoisia tai lihavia ja oireyhtymän kanssa yhtä aikaa esiintyykin useita, ainakin osittain ylipainosta johtuvia, metabolisia häiriöitä. Lukuisista tutkimuksista huolimatta on kuitenkin epäselvää, altistaako PCOS itsessään metabolisille häiriöille sekä sydän- ja verisuonisairauksille. Väitöskirjatutkimuksen tavoitteena oli selvittää, onko PCOS itsenäinen metabolisten ja sydän- ja verisuonisairauksien riskiä lisäävä tekijä. Tutkimus pohjautui Pohjois-Suomen syntymäkohortti 1966 tutkimuksen 14-, 31- ja 46-vuotisseurantoihin. PCOS luokittelu perustui 31- ja 46-vuotiskyselyissä itse ilmoitettuihin tyypillisiin PCOS oireisiin ja/tai diagnoosiin. Tutkimuksessa havaittiin, että 14- ja 31-ikävuoden välillä tapahtuva painonnousu oli yhteydessä PCOS diagnoosiin myöhemmällä iällä. 46-vuotiaana normaalipainoisilla PCOS naisilla ei ollut suurentunut tyypin 2 diabetes riski, mutta painonnousu varhaisaikuisuudessa oli merkittävästi yhteydessä sokeriaineenvaihdunnan häiriöön PCOS naisilla. PCOS oli yhteydessä kohonneeseen verenpaineeseen 31-vuotiaana ja hypertensioon 46-vuotiaana ylipainosta riippumatta. Oireyhtymään liittyvät metaboliset häiriöt olivat tärkein sydämen autonomisen hermoston säätelyyn vaikuttava tekijää, kun taas PCOS itsessään ei vaikuttanut autonomisen hermoston toimintaan. PCOS:ään sairastavien naisten sydämen rakenne ja funktio eivät merkitsevästi poikenneet kontrolloiden vastaavista muuttujista. Kuitenkin suhteellisen nuoresta iästä huolimatta PCOS naisilla esiintyi enemmän sydäninfarkteja ja kaksi kertaa enemmän sydän- ja verisuonitapahtumia, kuin kontrolleilla. Tutkimuksen tulokset osoittavat, että vaikkakin PCOS on itsenäinen riskitekijä metabolisille häiriöille, oireyhtymään liittyvä ylipaino vaikuttaa merkittävästi metabolisten häiriöiden esiintymiseen. PCOS:n ja sydän- ja verisuonitautitapahtumien yhteyden tarkempi tutkiminen vaatii kohortin jatkoseurantaa. Painonhallinnan tukemisen tulisi olla PCOS:ää sairastavien naisten hoidon kulmakivi. Asiasanat: aineenvaihdunta, autonominen hermosto, hyperandrogenismi, munasarjojen monirakkulatauti, painoindeksi, prediabetes, sydän- ja verisuonitaudit, tyypin 2 diabetes, verenpainetauti

8

9 To my family

10 8

11 Acknowledgements The present study was carried out at the Department of Obstetrics and Gynaecology, PEDEGO research unit, Oulu University Hospital, Medical Research Center Oulu and the Faculty of Medicine, University of Oulu in The study is based on data from the Northern Finland Birth Cohort 1966 and I wish to honor and thank the founders and developers of this unique cohort data, Professor Paula Rantakallio, Professor Marjo-Riitta Järvelin and Professor Anna-Liisa Hartikainen. It has been a privilege to work with this excellent data. I wish to express my deepest gratitude and respect to Docent Laure Morin- Papunen for all the unconditional support, time and guidance during these years. I have been most privileged to work with her and to get to know her. In the same way, I am greatly thankful to Docent Terhi Piltonen for all the energetic and inspiring guidance, support and help during these years. I wish to thank Professor Juha Tapanainen, who has always shared his great knowledge and found time to help. I also wish to thank Professor Stephen Franks for important constructive criticism, which has improved the manuscripts. I wish to thank the reviewers of this thesis, Professor Risto Kaaja and Docent Oskari Heikinheimo. I also want to express warm thanks to my follow-up group members Docent Tytti Raudaskoski, Professor Ulla Puistola and Docent Marja Ojaniemi for their warm support and constructive comments. I wish to thank our important collaborators Professor Sirkka Keinänen-Kiukaanniemi, PhD Juha Auvinen, PhD Kari Kaikkonen and PhD Antti Kiviniemi. I am also greatly thankful to PhD Tanja Nordström and MSc Jari Jokelainen who have patiently helped me with the cohort data. Special thanks go to PhD Sammeli West for working with me in these projects. I would also like to thank post-graduates Pekka Pinola and Johanna Puurunen, and peer PhD students Marika Kangasniemi, Salla Karjula, Maria-Elina Mosorin, Johanna Laru and Masuma Khatum for many memorable moments and the great atmosphere in our group. I also want to thank all my dear friends and my family for priceless support and never-ending kindness. Thank you for being in my life. This study was financially supported by Oulu University Hospital, the Finnish Medical Foundation, the 1.3milj club and the Paulo Foundation, all of which are gratefully acknowledged. Oulu, April 2019 Meri-Maija Ollila 9

12 10

13 Abbreviations AGM Abnormal glucose metabolism AMH Anti-Müllerian hormone AMI Acute myocardial infarction ART Assisted reproductive technology ASRM American Society of Reproductive Medicine BMI Body mass index BP Blood pressure BRS Baroreflex sensitivity cft Calculated free testosterone CVD Cardiovascular disease DHEA Dehydroepiandrosterone DHEAS Dehydroepiandrosterone Sulphate E Peak velocity of early diastolic transmitral flow e Peak velocity of early diastolic mitral annulus motion ECG Electrocardiogram EF Ejection fraction ESHRE European Society of Human Reproduction and Embryology FAI Free androgen index GnRH Gonadotrophin releasing hormone HA Hyperandrogenism HbA1c Glycated haemoglobin HDL High density lipoprotein HF High frequency HOMA-IR Homeostasis model assessment of insulin resistance HR Heart rate HRV Heart rate variability hscrp High-sensitivity C-reactive protein HTA Hypertension ICD International Classification of Diseases IR Insulin resistance IFG Impaired fasting glucose IGT Impaired glucose tolerance IVSd Interventricular septal end diastole LAESV Left atrial end systolic volume LC-MS/MS Liquid chromatography-tandem mass spectometry 11

14 LF Low frequency LH Luteinizing hormone LV Left ventricle/ventricular LVIDd Left ventricular internal diameter end diastole LVMI Left ventricular mass index MSNA Muscle sympathetic nerve activity nagm New abnormal glucose metabolism NFBC66 Northern Finland Birth Cohort 1966 NGT Normal glucose tolerance NIH National Institutes of Health nt2dm New T2DM nu Normalised unit OA Oligo-anovulation OGTT Oral glucose tolerance test OR Odds ratio PCOM Polycystic ovarian morphology PCOS Polycystic ovary syndrome pnn50 Percentage of successive differences in RRi > 50 ms Pre-DM Pre-diabetes pt2dm Previously diagnosed T2DM QUICKI Quantitative insulin sensitivity check index RAS Renin-angiotensin system RRi R-R interval RMSSD Root mean square of successive R-R differences SDANN Standard deviation of average normal-to-normal intervals SDNN Standard deviation of all RRi(s) SHBG Sex hormone-binding globulin SNS Sympathetic nervous system T Testosterone T2DM Type 2 diabetes mellitus ULF Ultra-low frequency VLF Very low frequency WHO World Heath Organisation WHR Waist-to-hip ratio 12

15 List of original publications This thesis is based on the following publications, which are referred to throughout the text by their Roman numerals: I Ollila, MM., Piltonen, T., Puukka, K., Ruokonen, A., Järvelin, M., Tapanainen, JS., Franks, S., Morin-Papunen, L. (2016). Weight Gain and Dyslipidemia in Early Adulthood Associate With Polycystic Ovary Syndrome: Prospective Cohort Study. The Journal of Clinical Endocrinology & Metabolism, 101(2), doi://dx.doi.org/ /jc II Ollila, MM., West, S., Keinänen-Kiukaaniemi, S., Jokelainen, J., Auvinen, J., Puukka, K., Ruokonen, A., Järvelin, MR., Tapanainen, JS., Franks, S., Piltonen, TT., Morin- Papunen, L. (2017). Overweight and obese but not normal weight women with PCOS are at increased risk of Type 2 diabetes mellitus a prospective population-based cohort study. Human Reproduction, 32(2), doi: /humrep/dew329 III Ollila, MM., Kaikkonen, K., Järvelin, M., Huikuri, HV., Tapanainen, JS., Franks, S., Piltonen TT., Morin-Papunen, L. (2019). Self-reported Polycystic Ovary Syndrome is Associated with Hypertension: A Northern Finland Birth Cohort 1966 Study. The Journal of Clinical Endocrinology and Metabolism, 104(4), doi: /jc IV Ollila, MM., Kiviniemi, A., Stener-Victorin, E., Tulppo, M., Puukka, K., Ruokonen, A., Tapanainen, JS., Franks, S., Morin-Papunen L., Piltonen, TT. Cardiac Autonomic Balance at age 46 in Women with PCOS - A Cohort Study of 1856 women. Manuscript. 13

16 14

17 Contents Abstract Tiivistelmä Acknowledgements 9 Abbreviations 11 List of original publications 13 Contents 15 1 Introduction 17 2 Review of the literature Polycystic ovary syndrome Definition and prevalence Aetiology of PCOS Clinical features Overweight and obesity Classification of overweight and obesity Overweight and obesity in PCOS Hormonal and metabolic effects of overweight and obesity Glucose metabolism in PCOS Classification of glucose metabolism Insulin resistance in PCOS Abnormal glucose metabolism in PCOS Blood pressure in PCOS Definition of hypertension PCOS and hypertension Effect of blood pressure on cardiac structure and function Cardiac autonomic function in PCOS Assessment of cardiac autonomic function Autonomic function in women with PCOS Cardiovascular diseases in PCOS Purpose of the present study 47 4 Study subjects and methods Study population Definition of PCOS and control groups Methods Statistical analysis

18 5 Results and Discussion Association of weight gain with the development of PCOS by age 46 (Study I) Abnormal glucose metabolism in women with PCOS (Study II) Hypertension and cardiac structure and function in PCOS (Study III) PCOS and cardiac autonomic function (Study IV) PCOS and cardiovascular disease morbidity (Study III) Strengths and limitations Conclusions 81 References 83 Original publications

19 1 Introduction Polycystic ovary syndrome (PCOS) presents with widely ranging health implications in women, such as reproductive, metabolic, and psychological issues. The syndrome affects 8 to 13% of the reproductive aged female population and is thus the most common endocrine disorder of fertile aged women (Teede et al., 2018). In the United States, the estimated annual total cost of evaluating and providing care to women with PCOS was approximately 4.36 billion U.S. dollars in the early 2000s (Azziz, Marin, Hoq, Badamgarav, & Song, 2005). According to Rotterdam criteria, PCOS is defined as the presence of two of the following three features: polycystic ovarian morphology in ultrasonography, oligoand/or anovulation, or biochemical or clinical hyperandrogenism (Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group, 2004; Teede et al., 2018). The exact aetiology of PCOS is unknown but it is thought to be a multifactorial disorder stemming from intrinsic/genetic and environmental predispositions. The majority of women with PCOS are overweight or obese, but it has remained unclear whether weight gain and high body mass index (BMI) predispose women to the development of PCOS or whether PCOS per se predisposes them to obesity (Hoeger & Oberfield, 2012). Insulin resistance (IR) often coincides with PCOS independently of BMI, exposing women with PCOS to glucose metabolism disorders (pre-diabetes and type 2 diabetes mellitus [T2DM]) (Diamanti-Kandarakis & Papavassiliou, 2006). Several studies have shown that PCOS is a risk factor of T2DM independently of BMI (Moran, Misso, Wild, & Norman, 2010), but it is unclear whether normal-weight women with PCOS are at an increased risk of T2DM. Similarly, women with PCOS often present with elevated blood pressure (BP) (Joham, Boyle, Zoungas, & Teede, 2015), but it has not been elucidated whether PCOS is an independent risk factor of hypertension, or if women with PCOS present with elevated BP as a result of confounding factors such as obesity and hyperandrogenism. Metabolic abnormalities, such as obesity, dyslipidaemia and hypertension are also often associated with impaired cardiac autonomic function; over-activity of the sympathetic autonomic nervous system and decreased parasympathetic activity. These derangements have also been shown to be risk factors of cardiovascular morbidity (Wulsin, Horn, Perry, Massaro, & D'Agostino, 2015). The results of previous studies have also suggested impaired cardiac autonomic balance in women with PCOS (Gui & Wang, 2017), although again, it is unclear if this is due to PCOS per se or due to other co-existing metabolic abnormalities. 17

20 As many studies have shown that women with PCOS display a clustering of cardiovascular disease risk factors (such as obesity, hyperandrogenism, abnormal glucose metabolism, dyslipidaemia, hypertension, and altered cardiac autonomic function), it has been postulated that such women might also have an increased risk of cardiovascular disease events, such as acute myocardial infarction and stroke, but the evidence is still controversial and further studies are awaited (Dokras, 2013). 18

21 2 Review of the literature 2.1 Polycystic ovary syndrome Definition and prevalence Polycystic ovary syndrome (PCOS) was first described by Chereau in 1844 (Chereau, 1844) and later by Stein and Leventhal in 1935 (Stein & Leventhal, 1935). Since then, the syndrome has been defined by various criteria. In 1990, the National Institutes of Health (NIH) established the first diagnostic criteria for PCOS: the presence of hyperandrogenism (HA) and chronic oligo-anovulation (OA), after exclusion of other diseases such as congenital adrenal hyperplasia, hyperprolactinaemia and androgen-secreting neoplasms (Zawadski & Dunaif, 1992). Since then, the combination of OA+HA has been referred to as classic PCOS. The diagnostic criteria for PCOS were updated in 2003 in Rotterdam by the European Society of Human Reproduction and Embryology/The American Society for Reproductive Medicine (ESHRE/ASRM) expert conference, which defined PCOS according to the presence of at least two of the following three features: menstrual irregularities (oligo- or anovulation), HA (clinical and/or biochemical) or polycystic ovarian morphology (PCOM), after excluding other aetiologies. The Rotterdam PCOS diagnostic criteria expanded the PCOS definition by adding two new PCOS phenotypes: PCOM+HA and PCOM+OA (Rotterdam ESHRE/ASRM- Sponsored PCOS Consensus Workshop Group, 2004). In 2006, The Androgen Excess and PCOS (AE-PCOS) Society published a third recommendation for the diagnosis of PCOS (Azziz et al., 2006). This recommendation included the presence of hyperandrogenism (hirsutism or hyperandrogenaemia) as a mandatory criterion for PCOS. Other features for PCOS diagnosis in that recommendation were the presence of ovarian dysfunction (OA or PCOM) and exclusion of other androgen excess or related disorders, such as 21- hydroxylase-deficient nonclassic adrenal hyperplasia, androgen-secreting neoplasms, androgenic/anabolic drug use or abuse, Cushing s syndrome, and the syndromes of severe insulin resistance, thyroid dysfunction, and hyperprolactinaemia. The aforementioned three recommendations for the diagnosis of PCOS are described in Table 1. 19

22 Table 1. Different diagnostic criteria for PCOS. Criteria Hyperandrogenism Oligo-anovulation PCOM NIH Rotterdam AE-PCOS Society PCOM: Polycystic ovarian morphology, NIH: National Institutes of Heath, AE: Androgen Excess Recently, in 2018, the first international evidence-based guidelines for the assessment and management of PCOS were published, including updated recommendations for the diagnosis of PCOS (Teede et al., 2018). The guidelines endorsed the Rotterdam criteria for PCOS diagnosis, i.e. the presence of two of the following criteria: OA, HA, or PCOM. In addition, the guidelines stressed that, in line with Rotterdam criteria, in the presence of both OA and HA, ultrasonographic assessment of the ovaries was not necessary for diagnosis, and that when using ultrasonography, different criteria for PCOM (Figure 1) should be used when using older and newer ultrasonographic technologies (Teede et al., 2018). Fig. 1. PCOM in ultrasonography. Copyright Laure Morin-Papunen. 20

23 In adolescence and during puberty many features of PCOS (irregular menstruation, acne, PCOM in ultrasonography) overlap with normal features, making PCOS diagnosis in adolescence challenging. Of note, ultrasonographic assessment of the ovaries should not be performed for those with a gynaecological age of less than eight years (Teede et al., 2018). However, as early as in puberty overweight and obese girls demonstrate consistent hyperandrogenaemia and hyperinsulinaemia compared with their normal-weight peers (McCartney et al., 2007), which might predict the development of PCOS (Nader, 2013). Moreover, in a previous study involving the Northern Finland Birth Cohort 1986 it was found that menstrual irregularities at age 16 were a good marker of hyperandrogenaemia, and that there was a significant linear trend towards higher free androgen index (FAI) values in the higher BMI quartiles (Pinola et al., 2012). Another study concerning the same cohort revealed that irregular menstruation and hyperandrogenaemia in adolescence were associated with PCOS and infertility in later life (West et al., 2014a). The generally accepted prevalence of PCOS is between 8 and 13% (Teede et al., 2018), but the prevalence greatly varies according to the diagnostic criteria used. Indeed, use of the Rotterdam criteria results in a three-fold increased prevalence compared with the prevalence when using NIH criteria (19.9% vs. 6.1%) (Yildiz, Bozdag, Yapici, Esinler, & Yarali, 2012). In addition, population characteristics such as ethnicity affect the prevalence of PCOS. A systematic review showed that the prevalence of PCOS was lowest among Chinese women, intermediate among Caucasian women and highest among Black women (Ding et al., 2017) Aetiology of PCOS The aetiology of PCOS remains under debate, but it is thought to be a complex, multifactorial developmental condition. Previous studies have suggested that PCOS could be a consequence of exposure to androgen excess in the intrauterine environment, which could lead to altered programming of the hypothalamicpituitary-ovarian axis, with increased gonadotrophin pulsatility and subsequent excess of androgen synthesis (Abbott, Dumesic, & Franks, 2002). The resulting phenotype is suggested to be further modified by environmental and genetic factors, which would explain the heterogeneous nature of the syndrome. PCOS clusters in families and although genome-wide association studies have identified a number of candidate regions, their role in contributing to PCOS is still largely unknown (Dumesic et al., 2015). 21

24 Recently, new data has also emerged on early-life growth and the prevalence of PCOS. Reduced foetal growth followed by infantile catch-up growth has been suggested to predispose girls to PCOS (de Zegher & Ibáñez, 2006) as well as early childhood weight gain (Koivuaho et al., 2019). Moreover, it has been proposed that normal-weight women with PCOS display the most severe defects in ovarian steroidogenesis, as they do not require any additional factors to develop the syndrome, whereas women with mild defects would need the contribution of other factors, such as obesity and IR, to develop the syndrome (Escobar-Morreale & San Millan, 2007; Homburg, 2009). On the other hand, the different phenotypes together with BMI manifestations may also represent different aetiologies behind the syndrome. Fig. 2. Roles of androgen excess and triggering factors during life for the development of PCOS. Recently, the role of anti-müllerian hormone (AMH), a glycoprotein secreted by the ovarian granulosa cells, has been under intense investigation in the aetiology of PCOS. Interestingly, it was found that in mice, AMH was able to modulate luteinizing hormone (LH) pulsatility, resulting in a PCOS-like phenotype characterized as ovulatory dysfunction and HA (Cimino et al., 2016). Given that pregnant women with PCOS show significantly higher AMH levels than controls, it has been postulated that AMH could possibly trigger HA during pregnancy (Tata et al., 2018). Indeed, the same group showed that in mice, elevated AMH levels were able to result in testosterone (T) excess in pregnant animals and consequent PCOS with reproductive and neuroendocrine phenotypes in the offspring (Tata et al., 2018). This publication was ground-breaking, as it was the first work to introduce a mechanism for PCOS aetiology. However, the causal relationship between AMH levels and development of PCOS is still under debate. 22

25 2.1.3 Clinical features Women with PCOS usually seek medical help for hirsutism, irregular menstruation or infertility. In addition to these main clinical features, 50 to 80% of women with PCOS are overweight or obese, and the majority of them present with IR (Ovalle & Azziz, 2002). Hyperandrogenism Hyperandrogenism can be either clinical (i.e. expressed as hirsutism) or biochemical. Hirsutism, excessive male-type terminal hair growth, is a common clinical manifestation of HA, together with acne, and less commonly, femalepattern hair loss. Of note, acne and female-pattern hair loss are not diagnostic features of HA. The clinical evaluation of hirsutism is based on visual inspection of nine skin areas (scores 1 to 4) and hirsutism is defined as a modified Ferriman Gallwey score of 4 6, depending on ethnicity (Ferriman & Gallwey, 1961; Teede et al., 2018; Yildiz, Bolour, Woods, Moore, & Azziz, 2010). Fig. 3. Modified Ferriman & Gallwey scoring sheet for the clinical evaluation of hirsutism. Published by permission of Elsevier. Biochemical HA can be defined as elevated levels of total serum T, calculated free testosterone (cft), FAI, androstenedione, or dehydroepiandrosterone sulphate 23

26 (DHEAS). The evidence-based guideline recommends that in PCOS, hyperandrogenaemia should be assessed using cft, FAI or calculated bioavailable T (Teede et al., 2018). Liquid chromatography-mass spectrometry (LC MS/MS) and extraction/chromatography immunoassay are the most accurate assessment methods for total or free testosterone in PCOS (Teede et al., 2018). However, the gold standard techniques for T measurement are mass spectrometry and equilibrium analysis, of which LC MS/MS is the method of choice for routine steroid hormone determination in the clinical laboratory (Taylor, Keevil, & Huhtaniemi, 2015). The FAI is widely used, but its role as an indicator of hyperandrogenaemia, especially in women with PCOS, is affected by IR, as IR is associated with lower circulating concentrations of sex hormone-binding globulin (SHBG) and thus increased FAI values (Wallace, McKinley, Bell, & Hunter, 2013). Calculated FT correlates well with free T concentrations measured by equilibrium dialysis (Vermeulen, Verdonck, & Kaufman, 1999). The origin of HA in PCOS is complex. In women with PCOS, gonadotrophinreleasing hormone (GnRH) pulse frequency from the hypothalamus is increased, possibly as a consequence of neuroendocrine impairment related to kisspeptin and GABA neurons (Katulski, Podfigurna, Czyzyk, Meczekalski, & Genazzani, 2018; Moore & Campbell, 2016). Increased GnRH pulse frequency leads to increased LH pulse frequency and amplitude in the anterior pituitary and thus increased serum LH concentrations (Blank, McCartney, Helm, & Marshall, 2007). This leads to increased T secretion in the ovarian theca cells, which are under LH regulation. The GnRH pulse generator is relatively resistant to negative feedback of progesterone and androgens, and an excess of circulating androgens does not have efficient negative feedback control in females (Nisenblat & Norman, 2009). Furthermore, both in vivo and in vitro studies have revealed that the ovarian theca cells of women with PCOS display intrinsic steroidogenic dysfunction, as they convert androgen precursors more effectively to T than normal theca cells, further enhancing the T excess (Rosenfield & Ehrmann, 2016). Furthermore, IR and compensatory hyperinsulinaemia, which commonly exist in women with PCOS, also contribute to hyperandrogenaemia. Insulin acts synergistically with LH, increasing T production from theca cells and inhibiting hepatic synthesis of SHBG, which leads to increased amounts of unbound, biologically active T (Ehrmann, 2005). In addition to ovarian androgen excess, 20 36% of women with PCOS also demonstrate adrenal androgen excess, the role and mechanism of which are still unclear (Nisenblat & Norman, 2009). 24

27 Fig. 4. Effects of insulin and LH on hyperandrogenaemia. Menstrual irregularities and reproductive disturbances Chronic anovulation often manifests as oligoamenorrhoea (menstrual cycles of > 35 days or < 8 cycles per year) or amenorrhoea (Ehrmann, 2005; Teede et al., 2018). The main factors leading to anovulation in women with PCOS are suggested to be associated with the accumulation of small follicles in the polycystic ovary (Homburg, 2009). HA contributes to this by accelerating the progression of preantral and small antral follicles from their precursors (Homburg, 2009), but on the other hand the small follicles are not recruited forward, resulting in disrupted folliculogenesis, and anovulation (Franks, Stark, & Hardy, 2008). Anovulatory cycles may lead to dysfunctional uterine bleeding and decreased fertility, but they also cause progesterone deficiency in the endometrium (Piltonen, 2016). Moreover, women with PCOS display an increased risk of endometrial cancer (Barry, Azizia, & Hardiman, 2014; Piltonen, 2016). PCOS is the most common cause of anovulatory infertility, and even though pregnancies occurring without assisted reproductive technology (ART) are relatively frequent, many women with PCOS need medical help to conceive (Joham, Teede, Ranasinha, Zoungas, & Boyle, 2015). On the other hand, women with PCOS have been reported to have at least one child as often as non-symptomatic women, although non-symptomatic women more commonly have two deliveries and larger family sizes (West et al., 2014b). Furthermore, women with PCOS may be at an increased risk of pregnancy complications such as pre-eclampsia, pregnancy- 25

28 induced hypertension and gestational diabetes (Palomba et al., 2015). A recent study, however, revealed that perinatal complications did not differ between PCOS and control women after adjusting for gestational weight gain (Kent et al., 2018). Life-long effects PCOS is a syndrome with life-long effects on women s health. The results of some studies have suggested that a predisposition to PCOS begins as early as in childhood, as children born small for gestational age have been reported to be at an increased risk of PCOS (Koivuaho et al., 2019). The cardinal features of PCOS become evident after puberty, and usually remain stable during reproductive years, when women with PCOS suffer from HA, irregular menstruation, infertility, obesity and metabolic as well as psychological problems (Karjula et al., 2017; Pinola et al., 2015; Teede et al., 2018). However, with aging the clinical symptoms of PCOS often fade (Brown et al., 2011). Both ovarian and adrenal androgen secretion decrease with time and usually older women with PCOS display lower levels of total and free T, androstenedione and DHEAS than younger PCOS women, even though T and androstenedione levels remain higher than in non-pcos women (Piltonen et al., 2004; Pinola et al., 2015; Welt & Carmina, 2013). Similarly, PCOM normalizes with aging as both ovarian volume and follicle numbers decrease (Koivunen et al., 1999; Piltonen et al., 2005; Welt & Carmina, 2013). As a result of this, women with PCOS may achieve regular menstrual cycles (Elting, Kwee, Korsen, Rekers-Mombarg, & Schoemaker, 2003), but whether this improvement results in a prolonged fertility period in women with PCOS compared with other women remains elusive. It has been reported that women with PCOS reach menopause significantly later than control women (Forslund et al., 2018; Li, Eriksson, Czene, Hall, & Rodriguez-Wallberg, 2016). The appearance of ovulatory cycles with aging also seems to impact on metabolic and cardiovascular disease risk factors, as a 20-year follow-up of hyperandrogenic women with PCOS showed that those recovering ovulatory cycles had less disturbed glucose and lipid metabolism than women with PCOS and anovulatory cycles (Carmina, Campagna, & Lobo, 2013). 26

29 Adolescence Obesity Hirsutism Irregular menses Acne Psychological stress? Reproductive age Obesity Hirsutism Metabolic abnormalities Infertility Pregnancy complications Psychological stress Pre and postmenopause Obesity Hirsutism Metabolic abnormalities Endometrial cancer Cardiovascular diseases? Psychological stress Fig. 5. The changing characteristics of PCOS during life. 2.2 Overweight and obesity Classification of overweight and obesity The prevalence of overweight and obesity is dramatically increasing worldwide, driven by western lifestyle, including reduced physical activity and excess food consumption. However, the rising prevalence seems to plateau in children and adolescents in high-income countries (Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in million children, adolescents, and adults. 2017). BMI and waist circumference are often used in clinical practice to evaluate overweight and obesity. BMI is a simple index of weight-to-height ratio (kg/m 2 ) that is commonly used to classify underweight (BMI < kg/m 2 ), normal weight (BMI kg/m 2 ), overweight (BMI kg/m 2 ) and obesity (BMI kg/m 2 ) in adults (Obesity: preventing and managing the global epidemic. Report of a WHO consultation. 2000). Waist circumference, measured at the level of the approximate midpoint between the lower margin of the last palpable rib and the top of the iliac crest, is also often used to estimate the distribution of excess adipose tissue. In females, waist circumference over 80 cm increases the risk of metabolic complications (World Health Organization, 2011) Overweight and obesity in PCOS The majority of women with PCOS are overweight or obese, although the prevalence of overweight and obesity is widely variable in reported series of cases. In the United States, as many as 74% of women with PCOS have been reported to be obese (Yildiz, Knochenhauer, & Azziz, 2008) and in an Italian study, more than 50% of women with PCOS were overweight or obese (Gambineri, Pelusi, Vicennati, 27

30 Pagotto, & Pasquali, 2002). In contrast, in a Greek study it was reported that BMI was similar in PCOS and reference groups (Diamanti-Kandarakis et al., 1999). Besides the major variation of the prevalence of overweight and obesity between different countries and ethnicities, these results also reflect differences in diagnostic criteria, PCOS phenotypes, and study populations. Moreover, it has remained unclear whether women are obese because of the syndrome or whether obesity predisposes women to the development of PCOS (Hoeger & Oberfield, 2012). Weight change during life in cases of PCOS Weight gain has been proposed to predispose women to the development of PCOS, as the condition frequently becomes manifest after significant weight gain (Escobar-Morreale & San Millan, 2007). On the other hand, weight reduction diminishes PCOS symptoms, as it reduces HA and IR as well as restoring ovulation (Escobar-Morreale, Botella-Carretero, A lvarez-blasco, Sancho, & San Millań, 2005). Interestingly, in a longitudinal general-population-based study from Australia three BMI trajectories (low-stable, moderately-rising and high-rising) were identified, and it was found that, compared with control women, women with PCOS were 1.6 times more likely to belong to the moderately-rising trajectory and 4.7 times more likely to belong to the high-rising trajectory (Kakoly, Earnest, Moran, Teede, & Joham, 2017). Another longitudinal general-population-based follow-up study revealed that obesity and greater weight gain between ages 20 and 30 were associated with the prevalence of PCOS at age (Teede et al., 2013). That study also showed that every BMI unit increase elevated the risk of PCOS by 9.2% (Teede et al., 2013) Hormonal and metabolic effects of overweight and obesity Adipose tissue, particularly visceral tissue, is an active metabolic and endocrine organ, secreting numerous molecules such as leptin, cytokines, adiponectin, complement components, and proteins of the renin-angiotensin system (Kershaw & Flier, 2004). It is also an important site for sex-steroid metabolism. Indeed, adipose tissue is an important site for peripheral conversion of androgen precursors (DHEA and DHEAS) to T, and its aromatase activity enables the conversion of androgens to oestrogens (Payne & Hales, 2004). Excess weight and adipose tissue are associated with IR and compensatory hyperinsulinaemia via increased secretion of proinflammatory cytokines, such as tumour necrosis factor alpha and leukotriene 28

31 B4, as well as free fatty acids, and worsening IR in muscle and liver tissues, further promoting hyperinsulinaemia. Furthermore, obesity significantly exacerbates all metabolic abnormalities, such as pre-diabetes (pre-dm), diabetes and metabolic syndrome, as well as the reproductive disturbances associated with the syndrome (Lim, Norman, Davies, & Moran, 2013). Some studies have suggested that women with PCOS more often display visceral/abdominal fat accumulation than control women (Carmina et al., 2007; Yildirim, Sabir, & Kaleli, 2003), whereas other studies have not shown any differences between women with PCOS and control women (Barber et al., 2008; Mannerås-Holm et al., 2011). Women with PCOS also have larger adipocytes, lower adiponectin levels, and lower adipose tissue lipoprotein-lipase activity when compared with BMI- and age-matched controls. Moreover, in a multivariate regression analysis, these adipose tissue-related disturbances played a more important role than hyperandrogenaemia to explain the presence of insulin resistance in women with PCOS (Mannerås-Holm et al., 2011). Fig. 6. Vicious cycle of metabolic disturbances in women with PCOS. 29

32 2.3 Glucose metabolism in PCOS Classification of glucose metabolism According to WHO classification, glucose metabolism can be classified as normal, pre-dm, including impaired fasting glucose (IFG) or impaired glucose tolerance (IGT), or T2DM (Table 2) (Alberti & Zimmet, 1998). Table 2. Classification of glucose metabolism based on 2-hour oral glucose tolerance test (2h-OGTT) results according to the WHO. Categories Fasting plasma glucose 2-hour oral glucose tolerance test (75 g) Normal glucose tolerance 6.0 mmol/l and < 7.8 mmol/l Impaired fasting glucose mmol/l and < 7.8 mmol/l Impaired glucose tolerance 6.0 mmol/l and mmol/l Type 2 diabetes mellitus 7.0 mmol/l or 11.0 mmol/l Glucose metabolism can be evaluated in various ways, such as assay of fasting plasma glucose, use of the 2h-OGTT, assay of glycated haemoglobin (HbA1c), performance of the frequently sampled intravenous glucose tolerance test, or the euglycaemic-hyperinsulinaemic clamp. Of these, the last two also give a measure of insulin sensitivity/resistance, as do many mathematical indexes, such as homeostasis model assessment for insulin resistance (HOMA-IR), the Matsuda index and the quantitative insulin sensitivity check index (QUICKI), which are based on fasting glucose and insulin or their values in the 2h-OGTT. The euglycaemic-hyperinsulinaemic clamp is considered to be the gold standard for assessment of IR, but it is a demanding, invasive and time-consuming method, and not feasible in everyday clinical practice. Instead, HOMA-IR and the 2h-OGTT are less time consuming and thus in wider use. However, the mathematical indexes (such as HOMA-IR, Matsuda, QUICKI etc.) are not fully accurate for the identification of IR; a large study of women with PCOS revealed that these surrogate indexes erroneously diagnose many subjects with PCOS as insulin sensitive, especially in the normal-weight group of women (Tosi, Bonora, & Moghetti, 2017). Moreover, HbA1c and fasting plasma glucose assays and the 2h- OGTT also have their limitations. Assay of HbA1c is neither sensitive nor sufficiently specific to detect pre-diabetes, fasting glucose is specific but not sensitive, and the 2h glucose values in OGTTs are so variable that the test should 30

33 always be performed twice before deciding on a diagnosis of pre-diabetes or T2DM (Barry et al., 2017). The detection of glucose metabolism abnormalities is of great importance, as T2DM is a major risk factor of cardiovascular diseases (Booth, Kapral, Fung, & Tu, 2006) Insulin resistance in PCOS Insulin resistance (IR) is a pathological condition in which the ability to transfer glucose from the circulation into the cells in response to activation of the insulin signalling pathway is reduced. This leads to compensatory hyperinsulinaemia, as the pancreas increases insulin production to compensate for reduced insulin action and glucose uptake. In women with PCOS, IR is due to an intrinsic, unique defect in post-receptor insulin signalling in its classical metabolic target tissues, muscle and adipose tissue (Diamanti-Kandarakis & Papavassiliou, 2006). However, as normal insulin signalling is maintained in the ovaries, the compensatory hyperinsulinaemia promotes hyperandrogenaemia by activating the steroidogenesis pathways in the ovaries (Diamanti-Kandarakis & Papavassiliou, 2006). Moreover, overweight and obesity, particularly visceral obesity, together with PCOS, synergistically further impair glucose tolerance and increase IR (Dunaif, Segal, Futterweit, & Dobrjansky, 1989), though IR also occurs in many normal-weight women with PCOS (Stepto et al., 2013). The prevalence of IR in women with PCOS varies between 50 to 95% (Ovalle & Azziz, 2002; Stepto et al., 2013) as a consequence of different IR assessment techniques, PCOS definitions and BMIs of the study populations. IR is a progressive condition that increases the risks of pre-dm, T2DM, and metabolic syndrome. In women with PCOS, lifestyle interventions and treatments reducing IR and hyperinsulinaemia, such as metformin treatment, significantly decrease weight and may improve hyperandrogenaemia, menstrual pattern and ovulation (Morin-Papunen, Koivunen, Ruokonen, & Martikainen, 1998; Naderpoor et al., 2015; Rosenfield & Ehrmann, 2016). 31

34 Fig. 7. Interplay between excess adipose tissue, IR and androgen excess. Modified from Escobar-Morreale & San Millan, Testosterone and pancreatic beta cell function Besides defective insulin-signalling in pancreatic beta cells, T excess also seems to play a role in the development of abnormal glucose metabolism in women with PCOS. Pancreatic beta cells exhibit androgen receptors and long-term androgen excess causes chronic activation of these receptors, leading to insulin hypersecretion and secondary beta cell failure in females (Xu, Morford, & Mauvais-Jarvis, 2019). Interestingly, prenatal androgen exposure could alter the development and, later on, the function of pancreatic beta cells (Mauvais-Jarvis, 2016). Accordingly, it has been found in murine and ovine models that maternal testosterone exposure alters the female offsprings beta cell function by causing basal hyperinsulinaemia, even in the absence of IR, and reduced insulin secretion 32

35 (Hogg, Wood, McNeilly, & Duncan, 2011; Roland, Nunemaker, Keller, & Moenter, 2010) Abnormal glucose metabolism in PCOS It is generally accepted that women with PCOS have an increased prevalence of glucose metabolism abnormalities. In most studies, the prevalence of T2DM in women with PCOS has been around 10% (Ehrmann, Barnes, Rosenfield, Cavaghan, & Imperial, 1999; Hudecova et al., 2011; Joham, Ranasinha, Zoungas, Moran, & Teede, 2014). In a meta-analysis it was concluded that PCOS is associated with an increased risk of pre-dm and T2DM, independently of BMI (Moran et al., 2010). Furthermore, in a more recent meta-analysis and systematic review including only good- or fair-quality studies it was found that women with PCOS (based on NIH criteria) had an increased prevalence of IGT and T2DM. Importantly, this prevalence differed by ethnicity and increased with obesity (Kakoly et al., 2018). Indeed, the results of several other previous studies have suggested that PCOS itself increases the risk of glucose metabolism abnormalities (Joham et al., 2014; Legro, Kunselman, Dodson, & Dunaif, 1999; Wang et al., 2011), but some studies have linked the risk mainly to overweight and obesity, but not to PCOS per se (Boudreaux, Talbott, Kip, Brooks, & Witchel, 2006; Espinos-Gomez, Corcoy, & Calaf, 2009; Gambineri et al., 2012). Follow-up studies have highlighted the importance of excess weight, and especially weight gain, in the development of abnormal glucose metabolism in women with PCOS. Norman et al. found that PCOS subjects converting from normal to abnormal glucose metabolism displayed greater BMI, abdominal obesity, and weight gain compared with those remaining normoglycaemic (Norman, Masters, Milner, Wang, & Davies, 2001) and Morgan et al. showed that a 1% increase of BMI increases the risk of T2DM by 2% (Morgan, Jenkins-Jones, Currie, & Rees, 2012). The recently published evidence-based guidelines for the assessment and management of PCOS (Teede et al., 2018) recommend the screening of glucose metabolism by way of fasting plasma glucose, HbA1c, or the 2h-OGTT, of which the last one is particularly recommended for high-risk individuals, i.e., women with overweight, a history of impaired fasting glucose or gestational diabetes, a family history of T2DM, hypertension or high-risk ethnicity. The detection of abnormal glucose metabolism is important to prevent the development of the morbidity and mortality associated with diabetes (Booth et al., 2006). 33

36 2.4 Blood pressure in PCOS Definition of hypertension Hypertension is a major global health issue affecting approximately 40% of adults worldwide (WHO Global status report on noncommunicable diseases. 2011). The definitions of elevated BP are variable, but according to guidelines from the European Society of Hypertension (ESH) and the European Society of Cardiology (ECS), hypertension is defined as systolic BP 140 mmhg and/or diastolic BP 90 mmhg (Williams et al., 2018). Elevated BP levels can damage blood vessels and cause adverse changes in important target organs such as the heart, the kidneys and the brain. In a meta-analysis of 61 prospective studies it was concluded that in the general population increased BP levels are strongly and directly associated with vascular and overall mortality (Lewington, Clarke, Qizilbash, Peto, & Collins, 2002) PCOS and hypertension The prevalence of hypertension has been reported to be higher in women with PCOS compared with controls (Joham et al., 2015; Schmidt, Landin-Wilhelmsen, Brannstrom, & Dahlgren, 2011). The results of some studies have suggested that this is mainly due to excess weight (Luque-Ramirez, Alvarez-Blasco, Mendieta- Azcona, Botella-Carretero, & Escobar-Morreale, 2007), although others have suggested that PCOS per se increases the prevalence of hypertension (Zachurzok- Buczynska, Szydlowski, Gawlik, Wilk, & Malecka-Tendera, 2011) (Table 3). Blood pressure is under constant regulation. Previous studies have suggested that women with PCOS might display changes in the BP regulation system, predisposing them to hypertension. Activation of the renin-angiotensin system (RAS) has a major role in the pathophysiology of hypertension. Women with PCOS have been shown to display elevated renin levels, which were significantly correlated with androgen levels (Jaatinen et al., 1995). It has been reported that in a rat in vivo model, androgens might upregulate the RAS in proximal tubules in the kidneys, and thus increase the volume reabsorption rate and consequently blood pressure levels (Quan et al., 2004). Moreover, 3-month treatment of female rats with dihydrotestosterone resulted in elevation of BP and IR, and these changes were associated with significant upregulation of intrarenal angiotensinogen and the sympathetic nervous system, suggesting that hyperandrogenaemia could have an 34

37 effect on the RAS and BP level (Yanes et al., 2011), and also that it could have a direct effect on the sympathetic nervous system (Maranon et al., 2015). It could be speculated, based on the above-mentioned possible RAS activation in women with PCOS, that angiotensin-converting enzyme inhibitors might be an optimal choice for first-line antihypertensive medication in women with PCOS, but this issue needs to be further studied. Table 3. Previous studies concerning hypertension and blood pressure in women with PCOS. Study Number Age (years) Findings Zachurzok-Buczynska et al., 2011 N=43 Mean: 16 Obese groups: sig. 24-hour mean BP Elting, Korsen, Bezemer, & Schoemaker, 2001 N=346 Mean: 39 HTA prevalence: sig. Holte, Gennarelli, Berne, N=36 Mean: 26 Day-time systolic and mean Bergh, & Lithell, 1996 arterial BP sig. Joham et al., 2015 N=26 (with HTA) Range: Normal-weight group: HTA prevalence sig. ; Overweight/obese group: HTA prevalence Multivariate regression analysis: PCOS was not associated with HTA. Schmidt et al., 2011 N=35 Range: HTA prevalence sig. Office BP level: Luque-Ramirez et al., 2007 N=36 Mean: 24 HTA prevalence: Ambulatory or office BP levels: Chen et al., 2007 N=151 Mean: 24 FAI sig. associated with BP Meyer, McGrath, & Teede, N=100 Mean: 33 Ambulatory BP: 2005 Zimmermann et al., 1992 N=14 Range: 30 Ambulatory BP: HTA = hypertension, BP = Blood pressure, sig. = significantly, = higher, = comparable. Two groups have also reported that circulating levels of copeptin, a surrogate marker of antidiuretic hormone, are significantly higher in women with PCOS (Karbek et al., 2014; Taskin, Bulbul, Adali, Hismiogulları, & Inceboz, 2015). Furthermore, in the presence of IR, which often coexists with PCOS, the vasodilating effect of insulin can be lost (Eckel, Grundy, & Zimmet, 2005), promoting elevation of BP. Women with PCOS may also display changes in function of the autonomic nervous system (Saranya, Pal, Habeebullah, & Pal, 2014), 35

38 which is an important fast-acting regulator of BP. In addition, as the majority of women with PCOS are overweight/obese, they are also at a risk of obstructive sleep apnoea, which is a major risk factor of hypertension (Gonzaga, Bertolami, Bertolami, Amodeo, & Calhoun, 2015). However, despite all this, it has remained unclear whether PCOS per se is a risk factor of hypertension Effect of blood pressure on cardiac structure and function Long-lasting elevation of BP can cause changes in cardiac structure and function, as elevated BP increases the workload of the heart, leading to compensatory myocardial hypertrophy, impaired left ventricular (LV) relaxation, and increased left ventricle pressure, which in turn cause enlargement of the left atrium (Santos & Shah, 2014). Echocardiographic examination of the heart is easy to perform and widely available and generates several parameters that describe the heart s structure and function. Both LV ejection fraction (LVEF) and global strain describe the heart s systolic function. Of these, LVEF has been in clinical use for a long time, whereas global strain is a novel indicator of the overall systolic function of the LV and may reveal heart disease at an early stage, even when the LVEF is still normal (Smiseth, Torp, Opdahl, Haugaa, & Urheim, 2016). The diastolic function of the heart can also be impaired, which is usually due to impaired relaxation of the LV, reduced restoring forces, and/or increased LV chamber stiffness (Nagueh et al., 2016). The diastolic function of the heart can be evaluated by using several two-dimensional or Doppler parameters. According to a recent recommendation, the following parameters should be used for identifying diastolic dysfunction: left atrium volume index (describes cumulative effects of increased LV filling pressures over time), annular e velocity (septal e and lateral e ), average E/e ratio, and peak tricuspid regurgitation velocity (Nagueh et al., 2016). 36

39 Table 4. Echocardiographic parameters. Abbreviation Parameter s full name Systolic function LVEF Left ventricular ejection fraction Global strain Global longitudinal strain Cardiac structure IVSd Interventricular septal end diastole LVIDd Left ventricular internal diameter end diastole LVMI Left ventricular mass index Diastolic function LAESV Left atrial end systolic volume E Peak velocity of early diastolic transmitral flow e Peak velocity of early diastolic mitral annulus motion E/e Ratio of E/e TR Tricuspid regurgitation Cardiac structure and function in women with PCOS Only a few studies have concerned evaluation of cardiac structure and function in women with PCOS. A study of year old women recruited from an academic centre revealed that women with PCOS displayed higher LAESV and LVMI as well as lower LVEF and early to late mitral flow velocity ratio, suggesting impairment of both systolic and diastolic function in PCOS, although the role of hypertension was not assessed in that study (Orio et al., 2004). An American prospective cohort study of 42 women with PCOS in their 30s showed that PCOS was significantly associated with higher LVMI and LAESV compared with a reference population, even though women with PCOS had significantly lower systolic and diastolic BP (Wang et al., 2012). A Turkish study on 35 women with PCOS revealed that they displayed diastolic dysfunction (Tíras et al., 1999). In contrast, some studies have not shown any significant differences in echocardiographic parameters between PCOS and control women. Selcoki et al. performed echocardiographic examinations in 48 age- and BMI-matched Turkish women with PCOS and 21 healthy controls and found comparable left ventricular and atrial diameters, and similar ejection fractions as well as indices of diastolic function (Selcoki et al., 2010). In addition, a third Turkish study on echocardiographic features in 26 women with PCOS and 24 controls did not show differences in cardiac structure or function (Tekin et al., 2009). Interestingly, a case-control study of 150 women 37

40 divided into three groups according to PCOS and IR status (with both PCOS and IR, without PCOS and with IR, and without either PCOS or IR) revealed that LV function was impaired only in the groups with IR (Kosmala et al., 2008). 2.5 Cardiac autonomic function in PCOS The autonomic nervous system includes the sympathetic and parasympathetic nervous systems. As the parasympathetic nerve branches follow the vagus nerve (the 10 th cranial nerve) parasympathetic activity is often referred to as vagal activity. The parasympathetic nervous system is dominant in rest and digest situations whereas the sympathetic nervous system is active in so-called fight or flight situations. Sympathetic activity can be assessed by using many different methods, such as measurement of muscle sympathetic nerve activity (MSNA), plasma levels of noradrenaline and adrenaline, and noradrenaline spillover (Lansdown & Rees, 2012). Of these, MSNA and noradrenaline spillover are very accurate estimates of sympathetic activity, but they are also invasive and time demanding and thus not suitable for studying large populations. Noradrenaline spillover describes the sympathetic activity of a specific organ, whereas the plasma levels of noradrenaline and adrenaline describe the body s overall sympathetic tone. The function of the heart is under precise control of the autonomic nervous system, hormones, the surrounding temperature and mechanical stress from breathing. In normal situations in healthy individuals, there are small constant changes in the time intervals between consecutive heart beats (or the time interval between consecutive R peaks in the electrocardiogram, Figure 8), and this variation is called heart rate variability (HRV). 38

41 Fig. 8. An illustration of the RRi. When parasympathetic activity is increased (above), the pulse is slow and there are variations in the RRi, whereas when sympathetic activity is increased (below), the pulse is fast and variations in the RRi are small. Changes in autonomic activity are reflected in HRV so that parasympathetic activity increases it, whereas sympathetic activity decreases it. Moreover, high HRV is a marker of general good health, as it indicates that our bodies can react to changes in the internal and external environments. For example, respiratory arrhythmia, often seen in the electrocardiograms of young fit individuals is a consequence of the decrease in parasympathetic activity during inhalation leading to an increase in the heart rate during inhalation, whereas the opposite is observed during exhalation (Yasuma & Hayano, 2004). The clinical relevance of HRV was recognised in 1965, when it was found that when the foetal heart was exposed to distress, changes in HRV occurred before abnormalities in the heart rate itself (Hon & Lee, 1965). Later in life, decreased HRV has been associated with increased risks of many major global public-health problems, such as depression (Sgoifo, Carnevali, Alfonso, Maria de los Angeles Pico, & Amore, 2015), hypertension, diabetes, cardiovascular diseases and increased mortality (Wulsin et al., 2015) Assessment of cardiac autonomic function The function of the cardiac autonomic nervous system can be investigated noninvasively using HRV from electrocardiograms with the aid of mathematical methods, such as time-domain and frequency-domain methods (Table 5) (Heart rate variability: standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North 39

42 American Society of Pacing and Electrophysiology. 1996). Thus measurement of HRV is an optimal choice for assessment of cardiac autonomic function in large study populations. Table 5. Heart rate variability parameters. Parameter abbreviations Full name Modifiers/Describes Time-domain RMSSD Root mean square of successive If decreased HRV is R-R differences SDNN Standard deviation of all RRi(s) If decreased HRV is pnn50 Percentage of successive differences in RRi > 50 ms If decreased HRV (and vagal activity) is Frequency-domain ULF Ultra-low frequency (< Hz) Mainly modified by diurnal rhythm VLF Very low frequency ( Hz) Mainly modified by temperature and hormones (adrenaline, renin) LF Low frequency ( Hz) Describes mainly sympathetic activity HF High frequency ( Hz) Describes mainly parasympathetic activity LF/HF ratio Autonomic balance HRV = Heart rate variability. = decreased. Hz = Hertz. It is also important to note that besides surrounding temperature, hormone activity and the autonomic nervous system, many other factors also have an effect on HRV. Increasing age, BMI, and stress, as well as decreased physical activity, and the presence of chronic diseases (such as hypertension, coronary artery diseases, and diabetes), and acute diseases (such as myocardial infarction, and acute psychosis) decrease HRV (Heart rate variability: standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. 1996; Almeida- Santos et al., 2016; Föhr et al., 2016; Istenes et al., 2014; Valkonen-Korhonen et al., 2003). Women with PCOS have significantly higher leptin levels (Zheng, Du, & Li, 2017) and elevated leptin levels increase the activity of the sympathetic nervous system (Quilliot, Böhme, Zannad, & Ziegler, 2008), which might predispose women with PCOS to sympathetic overactivity and thus decreased HRV. The function of the autonomic nervous system is also related to inflammation, as higher levels of high-sensitivity C-reactive protein (hscrp) and interleukin-6 are 40

43 associated with impaired HRV (Haensel, Mills, Nelesen, Ziegler, & Dimsdale, 2008). This is important, as women with PCOS are often reported to have lowgrade inflammation (Spritzer, Lecke, Satler, & Morsch, 2015) Autonomic function in women with PCOS Women with PCOS typically present with many factors that are associated with autonomic dysfunction in the general population, such as weight excess, IR, hyperinsulinaemia, hypertension, dyslipidaemia, and obstructive sleep apnoea (Lambert, Straznicky, Lambert, Dixon, & Schlaich, 2010; Stuckey, Tulppo, Kiviniemi, & Petrella, 2014), and therefore could be at increased risk of impaired cardiac autonomic function. Previous studies have suggested that women with PCOS might display reduced parasympathetic (vagal) (Saranya et al., 2014; Tekin et al., 2008; Yildirir, Aybar, Kabakci, Yarali, & Oto, 2006) and increased sympathetic activity (Lambert et al., 2015). In a recent systematic review and metaanalysis including eight studies it was reported that women with PCOS (n = 243) might display cardiovascular autonomic dysfunction with reduced parasympathetic and increased sympathetic activity. Unfortunately, the meta-analysis did not cover assessment of the effect of confounding factors, even though women with PCOS had significantly higher BMIs, WHRs, as well as significantly higher levels of glucose, triglycerides, and total cholesterol compared with the control group (n = 211 women) (Gui & Wang, 2017). Previous studies among women with PCOS have involved the use of various methods, such as microneurography, measurement of sympathetic skin responses, heart rate variability, heart rate recovery and noradrenaline spillover measurement, and concerned only women with PCOS in their 20s and 30s (Table 6). In addition, the duration of HRV recordings has varied, which may bias the results, as HRV guidelines state that it is not appropriate to compare time-domain parameters derived from recordings of different durations (Heart rate variability: standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. 1996). 41

44 Table 6. Previous studies on autonomic function in women with PCOS. Study Design Study population Results Ji et al., 2018 Retrospective chart review Women with PCOS (n=35) had higher BMI SDNN and RMSSD ; LF, LFnu, and and BP than controls (n=32) LF/HF ratio ; HFnu Kilit & Paşalı Kilit, 2017 HRV 60 normal-weight women with PCOS and 60 SDNN, RMSSD, HF, LF, HFnu, LFnu, and age-matched controls LF/HF ratio Özkeçeci et al., h Holter Metabolically healthy PCOS (n=23) and SDNN, SDANN, and RMSSD controls (n=25) Lambert et al., 2015 Microneurography and HRV 19 overweight/obese PCOS, 21 Multiunit muscle SNS 1 activity ; HRV overweight/obese controls Hashim, Hamdan, & Al- Plasma epinephrine, SSK 2, HRV, 64 PCOS, 40 controls Sympathoexcitation more pronounced in Salihi, 2015 Valsalva ratio obese than in non-obese women with PCOS Saranya et al., 2014 HRV, HR and BP response to 31 PCOS, 30 controls Decreased parasympathetic and increased standing (30:15 ratio), deep sympathetic activity breathing (E:I ratio) Di Domenico et al., 2013 Rest and post mental-stress HRV 32 anovulatory classic PCOS (C-PCOS), 16 Rest-HRV. C-PCOS: mean RRi, PNN50 ovulatory PCOS, 23 controls and RMSSD during mental stress. Ovulatory-PCOS de Sá, Joceline Cássia Ferezini et al., 2011 HRV Women with PCOS (n=23) had higher weight, glucose, insulin, cholesterol and triglycerides SDNN, RMSSD, LF, and HF, which correlated with BMI than controls (n=23) 42

45 Study Design Study population Results Sverrisdóttir, Mogren, MSNA 3 20 PCOS, 18 controls MSNA burst frequency Kataoka, Janson, & Stener- Victorin, 2008 Yildirir et al., 2006 HRV 30 PCOS, 30 controls LFnu and LF/HF ratio, HF 1 SNS = Sympathetic nervous system, 2 SSK = Sympathetic skin response, 3 MSNA = Muscle sympathetic nerve activity 43

46 In conclusion, previous studies have been very heterogeneous regarding study populations and designs, and thus it remains unclear whether the autonomic dysfunction detected in women with PCOS is independent of overweight/obesity, IR, hyperandrogenaemia and metabolic abnormalities and whether the dysfunction persists during late reproductive years in affected women. 2.6 Cardiovascular diseases in PCOS Women with PCOS present with an increased prevalence of cardiovascular disease risk factors, but it has remained unclear whether they have increased morbidity and mortality due to cardiovascular diseases (Dokras, 2013). Previous studies have shown inconsistent findings regarding the risk of cardiovascular morbidity in women with PCOS, as some have shown an increased prevalence of events (Glintborg, Hass Rubin, Nybo, Abrahamsen, & Andersen, 2015; Glintborg, Rubin, Nybo, Abrahamsen, & Andersen, 2018; Mani et al., 2013), whereas others have not (Iftikhar et al., 2012; Lunde & Tanbo, 2007; Meun et al., 2018; Morgan et al., 2012; Schmidt et al., 2011), possibly because of variable study designs and populations. In addition, previous studies have concerned relatively young women when considering CVD events (Glintborg et al., 2018; Mani et al., 2013; Morgan et al., 2012). Table 7. Cardiovascular morbidity in women with PCOS in previous studies. Study Study design and Age of study Number of PCOS Results country of origin population (years) patients Meun et al., 2018 Glintborg et al., 2018 Merz et al., 2016 A prospective population-based cohort study from the Netherlands National registerbased study from Denmark Cross-sectional study from the United States Mean: 70 N=106 Incident CVD, Atherosclerosis Mean: 29 N= CVD event rate Mean: 62 N=25 CVD 44

47 Study Study design and Age of study Number of PCOS Results country of origin population (years) patients Glintborg et al., 2015 Nationwide register Mean: 31 N= Stroke and study from Denmark thrombosis Mani et al., yr retrospective cohort study from the UK Iftikhar et al., 2012 Retrospective cohort study from the United States Morgan et al., 2012 Retrospective observational study from the UK Schmidt et al., 2011 A 21-yr follow-up study from Sweden Lunde & Tanbo, Follow-up study 2007 from Norway Mean: 30 N=2301 Myocardial infarction Mean: 45 N=309 Myocardial infarction, coronary artery bypass graft surgery, stroke, death Mean: 27 N= Large vessel disease 1 Range: N=35 Myocardial infarction and stroke Range: N=149 CVD Lo et al., 2006 Register-based Mean: 31 N= CVD study from the United States Cibula et al., 2000 A cross-sectional study from the Range: N=28 Coronary artery disease Czech Republic Wild, Pierpoint, McKeigue, & Jacobs, 2000 Retrospective cohort study from the UK Mean: 57 (range: 38 98) N=319 Coronary artery disease, Cerebrovascular disease Pierpoint, McKeigue, Isaacs, Wild, & Jacobs, 1998 A 30-yr follow-up study from the UK Mean: 56 N=786 CVD Dahlgren, Janson, A prospective study Mean: 55 N=33 AMI or ischaemic Johansson, Lapidus, & Oden, 1992 from Sweden cerebral disease 1 Large vessel disease: myocardial infarction, stroke, angina, or central or peripheral revascularization. AMI: Acute myocardial infarction. 45

48 The overall risk of cardiovascular diseases is influenced by multiple factors, such as age, family history, lifestyle habits, and the presence of many metabolic changes. In clinical practice the role of individual risk factors and the prevention of future cardiovascular events must be evaluated in relation to the overall risk profile. Moreover, the independent role of hyperandrogenaemia as a metabolic risk factor is still controversial. A previous study revealed that women with PCOS and hyperandrogenaemia seemed to have a more adverse cardiometabolic profile and a greater number of CVD risk factors than non-hyperandrogenic women with PCOS, indicating that hyperandrogenaemia might play a role as an independent CVD risk factor (Daan et al., 2014). Interestingly, in the same study it was reported that this may not apply to women of Northern European descent. Moreover, the phenotype of PCOS also influenced the prevalence of CVD risk factors (Daan et al., 2014), and might therefore have an impact on the occurrence of later CVD events. However, in a recent study it was reported that postmenopausal women with high androgen levels and retrospectively diagnosed PCOS did not have an increased risk of CVD (Meun et al., 2018). It has been speculated that despite the high CVD risk profiles commonly seen in women with PCOS, later age at menopause and, consequently, more prolonged oestrogen exposure could alleviate the the CVD burden in PCOS. Later age at menopause (defined as menopause occurring at age 55 or later) was recently reported to be associated with significantly longer life expectancy compared with early-onset menopause (defined as menopause occurring at age 44 or earlier) in a general female population (Asllanaj et al., 2018). All in all, data on the risks of CVD events and mortality are still under debate. Whether these conflicting results are due to a lack of reliable data, variable study designs and/or to the fact that the women in most studies have been too young to display CVD events, remains to be resolved in future longitudinal studies. 46

49 3 Purpose of the present study The main purpose of the present study was to investigate the effects of PCOS on women s metabolic and cardiovascular health by using the Northern Finland Birth Cohort 1966 study population. The prospective, longitudinal study design, with follow-up from birth to late adulthood in this large cohort offered a unique opportunity to investigate the impact of PCOS on glucose metabolism and cardiometabolic morbidity. The study design also allowed us to investigate the effects of confounding factors, such as weight gain, weight excess, and lifestyle habits on the development of PCOS itself as well as on the development of abnormal glucose metabolism and adverse cardiometabolic outcomes. The specific aims were: 1. To investigate the association between weight gain during life and the development of PCOS, and to identify parameters associated with PCOS diagnosis by age 46 (Study I). 2. To study whether the prevalence of abnormal glucose metabolism is increased in women with PCOS by age 46, independently of BMI (Study II). 3. To explore whether women with PCOS have a higher risk of hypertension at the ages of 31 and 46, independently of BMI (Study III). 4. To gain insight into whether or not women with PCOS exhibit changes in the function of the cardiac autonomic nervous system (Study IV). 5. To investigate whether women with PCOS have an increased prevalence of cardiovascular morbidity and mortality (Study III). 47

50 48

51 4 Study subjects and methods 4.1 Study population The study is based on the Northern Finland Birth Cohort 1966 (NFBC1966), which is a large prospective general-population-based longitudinal birth cohort. All individuals with expected term in 1966 in the two northernmost provinces in Finland (Oulu and Lapland) were included in this birth cohort. The study population comprised all individuals born alive in 1966 (12,231 births, 5889 females). Enrolment in this database began at the 24 th gestational week and, so far, this cohort population has been followed at birth, and at ages 1, 14, 31 and 46. Postal questionnaires were sent at ages 14, 31 and 46, and clinical examinations were performed at ages 31 and 46. The study followed the principles of the Declaration of Helsinki. The Ethics Committee of the Northern Ostrobothnia Hospital District approved the research. All participants took part on a voluntary basis and signed an informed consent document. In 1980, at age 14, a postal questionnaire was sent to 5764 females and 95% of them (n = 5455) responded, with help from their parents. In 1997, at age 31, a comprehensive postal questionnaire was sent to 5608 women and 4523 (81%) of them responded. In addition, those living in Northern Finland or in the Helsinki metropolitan area (n = 4074 women) were invited to a clinical examination. Of these, 77% (n = 3127) participated in a clinical examination including anthropometric measurements and collection of blood samples for assessment of hormonal and metabolic parameters. Again, in 2012, at age 46, a comprehensive health investigation was carried out. A postal questionnaire and an invitation to a clinical examination were sent to all individuals alive and with known addresses (n = 5123 women). Of these, 3706 (72%) women responded to the questionnaire and 3280 women (64%) participated in the clinical examination. 49

52 14 years 31 years 46 years Postal questionnaire Postal questionnaire Postal questionnaire Weight and height Weight and height Waist circumference Weight and height Waist circumference Blood pressure Blood pressure Laboratory measurements Laboratory measurements Oral glucose tolerance test Echocardiography Heart rate variability Fig. 9. Data collection from the Northern Finland Birth Cohort 1966 included postal questionnaires and clinical examinations. 4.2 Definition of PCOS and control groups At age 31, the postal questionnaire included questions on oligo/amenorrhoea: Is your menstrual cycle often (more than twice a year) longer than 35 days? and excessive body hair: Do you have bothersome, excessive body hair growth? Of the women who responded to these questions, 4.1% (n = 125) reported both oligo/amenorrhoea and hirsutism (OA+H), after excluding pregnant women, women using hormonal preparations (n = 1459) or those not permitting the use of their data for analysis (n = 41). The validity of this questionnaire to distinguish PCOS cases has already been shown in our previous studies of the same cohort, as the women with both OA+H present the typical hormonal, metabolic and psychological characteristics of the syndrome (Karjula et al., 2017; Taponen et al., 2003; Taponen, Ahonkallio et al., 2004). At age 46, the postal questionnaire included the question: Have you ever been diagnosed as having polycystic ovaries and/or polycystic ovary syndrome (PCOS)?, to which 181 responded yes. Consequently, women who reported both OA+H at age 31 and/or diagnosis of PCOS by age 46 were classified as women with PCOS (n = 279). Women without 50

53 any PCOS symptoms at age 31 and without diagnosis of PCOS by age 46 were classified as controls (n = 1577). 4.3 Methods Anthropometrics At ages 14, 31 and 46 the study subjects reported their weight and height in the postal questionnaires. In addition, weight and height were measured by an experienced research nurse in the clinical examinations at ages 31 and 46. BMI was calculated as the ratio of weight (kg) to height squared (m 2 ). Women with BMI < 25.0 kg/m 2 were classified as normal weight and those with BMI 25.0 kg/m 2 as overweight/obese (Obesity: preventing and managing the global epidemic. Report of a WHO consultation. 2000). Waist circumference was measured at the level midway between the lowest rib margin and the iliac crest in the clinical examinations. Glucose metabolism Fasting plasma glucose and insulin were measured at ages 31 and 46. In addition, at age 46, a 2h-OGTT was performed in 2780 women after overnight (12-hour) fasting. Exclusion criteria for the OGTT were medication for diabetes or fasting capillary blood glucose level over 8.0 mmol/l. Both serum insulin and plasma glucose levels were measured at baseline and at 30, 60 and 120 minutes after a 75- g glucose intake. Glucose levels in fasting plasma (fp) and 2-hour OGTT samples at each time point were analysed and categorized according to WHO standards (Alberti & Zimmet, 1998): Normal glucose tolerance (NGT), IFG, IGT, Pre-DM: IFG or IGT, and T2DM. Furthermore, glucose metabolism disorders were classified more accurately: previously diagnosed T2DM (pt2dm), new T2DM diagnosis revealed by a 2h-OGTT at age 46 (nt2dm: fp-glucose 7.0 mmol/l and/or 2hglucose in a 2h-OGTT 11.1 mmol/l), all type-2 diabetes mellitus diagnoses (T2DM: pt2dm or nt2dm), abnormal glucose metabolism (AGM: Pre-DM or T2DM) and new abnormal glucose metabolism disturbance in a 2h-OGTT at age 46 (nagm: Pre-DM or nt2dm). All the pt2dm cases reported in postal questionnaires were verified and completed in hospital discharge and national drug registers from the Social Insurance Institution of Finland. 51

54 Table 8. Abbreviations and definitions of glucose metabolism status. Abbreviation Definition NGT Normal glucose tolerance IFG Impaired fasting glucose in 2h-OGTT IGT Impaired glucose tolerance in 2h-OGTT Pre-DM IFG or IGT pt2dm Previously diagnosed T2DM nt2dm New T2DM diagnosis revealed by 2h-OGTT at age 46 T2DM pt2dm or nt2dm AGM Abnormal glucose metabolism: pre-dm or T2DM nagm New AGM: pre-dm or nt2dm The classification of glucose metabolism is based on WHO criteria (Alberti & Zimmet, 1998). Blood pressure In the clinical examination, brachial systolic and diastolic BP were measured twice at age 31 and three times at age 46 (with a one-minute interval and after 15 minutes rest) from the right arm of the seated study subject, using a manual mercury sphygmomanometer at age 31 and an automated, oscillometric BP device with an appropriately sized cuff (Omron Digital Automatic Blood Pressure Monitor Model M10-IT) at age 46. Mean systolic and diastolic BP values were calculated. In addition, information concerning hypertension diagnosis and use of antihypertensive medication was gathered from postal questionnaires at ages 31 and 46. Cardiac structure and function At age 46, a sub-population was enrolled for echocardiographic examination at the Oulu research unit. As a rule, every second individual attending the clinical examination was randomly enrolled for echocardiographic examination without any preselection criteria, and without the participant s own wish influencing the enrolment. All participants agreed to take part in echocardiography. In total, echocardiography was performed in 645 women, including 37 women with PCOS and 283 control women. The transthoracic 2D echocardiography was performed on-line by an experienced cardiologist, using a General Electric Vivid E9 device with an M5S-D 1.5/4.6 MHz sector transducer for cardiovascular imaging (GE Health Medical, Horten, Norway). All measurements followed the American 52

55 Society of Echocardiography guidelines (Lang et al., 2005). In addition, global longitudinal strain, a novel indicator of the overall systolic function of the LV, was assessed off-line using EchoPac 7 software (automated function imaging) (Patrianakos et al., 2015). Cardiac autonomic function The clinical examination at age 46 included assessments of HRV and baroreflex sensitivity as part of the broad cardiovascular health status evaluation. The study subjects were informed about the measurement protocol (shown below, Figure 10). Fig. 10. Flow chart illustrating the HRV measurement protocol. A heart rate (HR) monitor (RS800CX, Polar Electro Oy, Kempele, Finland) was used to record RRi. In addition, standard lead-ii ECG (Cardiolife, Nihon Kohden, Tokyo, Japan), breathing frequency (MLT415/D, Nasal Temperature Probe, ADInstruments, Bella Vista, New South Wales, Australia), and BP by finger photoplethysmography (Nexfin, BMEYE Medical Systems, Amsterdam, the Netherlands) were recorded, with a sampling frequency of 1,000 Hz (PowerLab 8/35, ADInstruments). The preparations were followed by at least a 1-minute stabilization period before the beginning of the 3-minute recording period in a seated position. After that recording period, the participants stood up and remained still in a standing position for another three minutes while breathing normally. The first 150 seconds of recording in a seated position and the last 150 seconds in standing position were used in the analysis. The RRi data were edited on the basis 53

56 of visual inspection, artefacts and ectopic beats were removed and replaced with the local average (Hearts 1.2, University of Oulu, Oulu, Finland). Sequences with 10 consecutive beats of noise, and ectopic beats were deleted. RRi series with 80% accepted data were included in the analyses. The final study population included 1029 controls and 160 women with PCOS. Mean HR, RMSSD, spectral power densities (fast Fourier transformation, length 512 beats) including LF and HF components of HRV, and their ratio (LF/HF), were analysed. Spontaneous baroreflex sensitivity (BRS) was assessed in participants studied at the Oulu laboratory unit. Continuous ECG, BP and respiration signals were imported to custom-made stand-alone Matlab-based software (Biosignal Processing Team, University of Oulu, Oulu, Finland), where RRi and systolic BP values were extracted. Artefacts and ectopic beats were replaced using linear interpolation (< 5% for accepted recording) and thereafter, resampled at 2 Hz and detrended (< 0.04 Hz, Savitzky Golay method). A fast Fourier transform (Welch method, segments of 128 samples with 50% overlap) was performed to analyse low frequency (LF: Hz) power of RRi and systolic BP oscillations for subsequent analysis of BRS by the alpha method, if sufficient coherence ( 0.5) between LF oscillations in RRi and systolic BP was verified. BRS was successfully calculated in 609 controls and 105 women with PCOS. Cardiovascular morbidity and mortality In Finland, all individuals have a unique personal identification number. Individuals with serious acute illnesses are treated in public specialised health care centres, which report the diagnoses to the hospital discharge and hospital outpatient clinic registers. An individual s data can be retraced from these national registers using the unique personal identification number. ICD-8, ICD-9 and ICD-10 diagnostic codes for cardiovascular morbidity and events (I20, I21-25, I50, I60-I63, I63-I64, I65-I68, I69, I70, G45, G46) were identified from hospital discharge, hospital outpatient clinic and basic health care registers, with data covering , and , respectively. Cardiovascular mortality data was retrieved from the mortality register of Statistics Finland, which covers the information about cause and time of death. According to Finnish law, it is mandatory to report an occurrence of death and cause of death to this register and thus the coverage is practically 100%. 54

57 Laboratory methods At age 31, plasma glucose, and serum concentrations of total cholesterol, highdensity lipoprotein cholesterol (HDL-cholesterol), low-density lipoprotein cholesterol (LDL-cholesterol), triglycerides, insulin, hscrp and SHBG were all assayed as previously described (Taponen et al., 2003). Serum T concentrations at the ages of 31 and 46 were assayed using Agilent triple quadrupole 6410 LC/MS equipment (Agilent Technologies, Wilmington, DE, USA). The normal upper limit for T was 2.3 nmol/l at age 31 and 1.70 nmol/l at age 46, based on 97.5% percentiles calculated in this study population. At age 31 SHBG was assayed as previously described (Taponen et al., 2003), and at age 46 by chemiluminometric immunoassay (Immulite 2000, Siemens Healthcare, Llanberis, UK). As the SHBG analysis method had changed, serum SHBG concentrations analysed by the Immulite method were compared with concentrations analysed by the former method (fluoroimmunoassay, Wallac, Inc. Ltd., Turku, Finland) by using linear regression analysis (y = x ) and coefficient of determination (0.9915). Therefore, the SHBG values at age 31 were transformed to be comparable with the values analysed at age 46 by using the formula old method 31yr SHBG , and the results are reported accordingly. The FAI was calculated by using the equation 100 T (nmol/l)/shbg (nmol/l). Concentrations of HbA1c and total haemoglobin were measured by an immunochemical assay method (Advia 1800; Siemens Healthcare Diagnostics Inc., Tarrytown, Ny, USA) and their ratios were reported as mmol/mol. All samples were analysed at NordLab Oulu, a testing laboratory (T113) accredited by the Finnish Accreditation Service (FINAS) (EN ISO 15189) Statistical analysis BMI values at the ages of 31 and 46 were obtained mainly at the clinical examination and supplemented with self-reported BMI if clinically measured BMI was not available. There were no differences between the self-reported and clinically measured BMI values. Continuous data were presented as means ± standard deviation (SD) or as medians with 25% and 75% quartiles and differences in continuous parameters were analysed by Student s t-test or by the Mann Whitney U-test, as appropriate. Categorial data were analysed using crosstabulation and Pearson s Chi-squared (χ²) test or Fisher s exact test. Binary logistic 55

58 regression models as well as linear regression models were also used, with adjustment for appropriate confounding factors. Statistical analyses were performed using IBM SPSS Statistics 22.0 and 24.0 (SPSS, Inc., 1989, 2013, IBM Corp.). Values of p < 0.05 were considered as statistically significant. Specific methods in different studies Study I. Multivariate binary logistic regression models included the parameters significantly associated (p < 0.15) with the diagnosis of PCOS in the univariate regression analysis (Akaike, 1973). The FAI was used in the models as it is generally considered as a good indicator of biochemical hyperandrogenism in women with PCOS. HOMA-IR was selected to estimate insulin resistance, and serum levels of LDL and triglycerides were included in the models as they represent the most typical lipid disorders observed in women with PCOS. The maximum number of variables was limited to five in each model. Studies II and III. Binary logistic regression models were adjusted for consumption of alcohol (no use, light [> g/week], moderate [> g/week] and heavy use [> 210 g/week]), smoking (never a smoker, former smoker for > 6 months, former smoker for < 6 months and current smoker), education (basic, secondary, tertiary), current use of cholesterol-lowering drugs, and of combined contraceptive pills at age 46. In Study III effect sizes were reported as Cohen s d-values. Study IV. Women using beta-blockers were excluded from the HRV analysis (104 controls [7.7%] and 30 women with PCOS [13.3%]). Continuous variables with skewed distributions were transformed into natural logarithms (ln) for the analysis of difference between the PCOS and control groups. Mean arterial pressure (MAP) was calculated as follows: DBP + 1/3 (SBP - DBP). Correlation analysis as well as univariate and multivariate linear regression analyses were used to study factors associated with HRV parameters. First, univariate linear regression models were used to reveal the parameters significantly associated with the outcome variable. Then, step-wise multivariate models were used to identify the most important explanatory variables. The final multivariate model included the following as explanatory variables: PCOS, BMI, MAP, FAI, HOMA-IR and triglycerides. The number of explanatory variables included in the model had to be limited to avoid multicollinearity. BMI was selected for the model, as it significantly differs between PCOS and control women, and obesity is suggested to effect HRV. MAP was selected because it combines information from 56

59 both systolic and diastolic BP, and FAI because it is considered as a good indicator of hyperandrogenism in women with PCOS, and hyperandrogenism has been suggested to alter HRV in cases of PCOS. HOMA-IR was used as an estimate of insulin resistance. Anxiety was not included in the final multivariate model, because in the preliminary models using the step-wise method, anxiety was always the first variable to be excluded (indicating that anxiety was not significantly associated with RMSSD). The results of linear regression models were reported as unstandardized coefficients (B), the 95% confidence interval for B, p-values and the R 2 value for the model. The multicollinearity assumptions in the multivariate linear regression model were investigated using VIF, tolerance and eigenvalue indexes. In addition, histograms of regression standardized residual frequency, normal P-P plots of regression-standardized residuals and scatterplot figures were visually inspected to ensure that the model met the analysis assumption. 57

60 Table 9. Characteristics and results of Studies I IV. Study characteristics Study I Study II Study III Study IV Number of subjects Outcome measurement 125 women with PCOS symptoms (OA+H) at age 31, 181 women with self-reported PCOS diagnosis by age 46, 1577 controls Factors associated with symptoms or diagnosis of PCOS Main results BMI was significantly greater at ages 14, 31 and 46 in PCOS women. Weight gain in early adulthood was significantly associated with later PCOS diagnosis. 279 PCOS (i.e. OA+H at age 31 and/or selfreported 279 PCOS, 1577 controls 279 PCOS, 1577 controls PCOS diagnosis by age 46), 1577 controls Glucose metabolism abnormalities Hypertension, cardiac structure and function, CVD Normal-weight women with PCOS did not have increased odds for T2DM. Overweight/obese women with PCOS had significantly higher odds for T2DM than overweight/obese controls. PCOS was associated with hypertension, but BMI had higher odds for hypertension than PCOS. Prevalence of CVD was sig. higher in women with PCOS. Cardiac autonomic function Women with PCOS displayed impaired HRV, which was dependent on metabolic abnormalities. 58

61 5 Results and Discussion 5.1 Association of weight gain with the development of PCOS by age 46 (Study I) The aim of Study I was to clarify the change of BMI from age 14 to age 46 in women with PCOS compared with controls and to identify the factors associated with the presence of PCOS diagnosis by age 46. The most important finding was that women with typical PCOS symptoms (OA+H at age 31) or diagnosis of PCOS by age 46 had significantly greater BMI values (Figure 11) and greater prevalence of obesity compared with the control women at ages of 14, 31 and 46 years. Fig. 11. BMI values at ages 14, 31 and 46 in controls, in women with OA+H at age 31, and in women with self-reported PCOS diagnosis by age 46. ***p < and **p < In addition, women with self-reported symptoms at age 31 or self-reported diagnosis of PCOS by age 46 had experienced significantly greater weight gain in early adulthood (between ages 14 and 31) compared with control women, whereas weight gain in later adulthood (between ages 31 and 46) was comparable between these groups. Women with self-reported symptoms at age 31 or self-reported diagnosis of PCOS by age 46 also exhibited significantly greater abdominal obesity, measured by waist circumference, both at ages 31 and 46, compared with control women (Figure 12). 59

62 Fig. 12. Waist circumference at ages 31 and 46 in controls, in women with OA+H at age 31 and in women with self-reported PCOS diagnosis by age 46. ***p < In univariate binary logistic regression analysis, high BMI in adolescence, early adulthood and late adulthood, as well as waist circumference at ages 31 and 46, were significantly associated with self-reported diagnosis of PCOS by age 46 (Figure 13). Of these, BMI at age 31 was the strongest predictor (highest R 2 value 0.028). Moreover, the change in BMI between 14 and 31 years (R 2 value = 0.013), but not between 31 and 46 years, was a significant risk factor of PCOS. 60

63 Fig. 13. Univariate binary regression analysis for self-reported PCOS diagnosis by age 46, Part I. Additionally, at age 31, HOMA-IR, the FAI and serum concentrations of T, insulin, LDL, hscrp, total cholesterol and triglycerides were significantly positively associated and serum levels of SHBG negatively associated with a diagnosis of PCOS by age 46 in univariate regression analysis (Figure 14). 61

64 Fig. 14. Univariate binary regression analysis for self-reported diagnosis of PCOS by age 46, Part II. In multivariate regression analysis, the FAI (OR = 1.08, 95% CI: ) and serum levels of triglycerides (OR = 1.48, 95% CI: ), insulin (OR = 1.05, 95% CI: ), SHBG (OR = 1.01, 95% CI: ) and T (OR = 1.44, 95% CI: ) at age 31 were associated with PCOS by age 46, after adjustment for BMI at age 31. However, when BMI, the FAI, HOMA-IR and serum levels of LDL and triglycerides were included in the same multivariate binary regression analysis for explaining PCOS by age 46, only BMI at age 31 (OR = 1.05, 95% CI: ) and the FAI at age 31 (OR = 1.08, 95% CI: ) remained significantly associated with diagnosis of PCOS by age 46. The present data analysis completed and strengthened previous findings in this same cohort (Laitinen et al., 2003) by extending follow-up until late adulthood. The results showed that weight gain in early adulthood, but not later, was significantly associated with self-reported diagnosis of PCOS by age 46. Our results are also in line with those of an Australian longitudinal general-population-based study (Teede et al., 2013), showing that obesity and greater weight gain between the ages of 20 and 30 were associated with PCOS status at age Similarly, in a longitudinal general-population-based study it was also reported that women with PCOS were 1.6 times more likely to experience a moderate increase of weight and were

65 times more likely to experience a large increase of weight than healthy control women (Kakoly et al., 2017). In the light of a previous developmental hypothesis of the pathophysiology of PCOS, these results suggest strongly that major weight gain in early adulthood as well as and obesity are among the major triggering factors for the development of overt PCOS (Abbott et al., 2002; Escobar-Morreale & San Millan, 2007; Homburg, 2009). In line with this, an early childhood BMI rise has been associated with PCOS diagnosis (Koivuaho et al., 2019). A high prevalence of overweight and obesity might also predispose women with PCOS to cardiovascular diseases and a shorter life span. Accordingly, in a large cohort study from the United Kingdom it was reported that individuals of normal weight had a 3.5-year longer life expectancy compared with obese women, and that BMI had a J-shaped association with cardiovascular diseases independently of confounding factors (Bhaskaran, Dos-Santos-Silva, Leon, Douglas, & Smeeth, 2018). 5.2 Abnormal glucose metabolism in women with PCOS (Study II) Study II was focused on glucose metabolism abnormalities in women with PCOS and we aimed to investigate the respective relationships between PCOS, long-term weight gain and obesity with the development of pre-dm and T2DM by age 46. The main finding was that only overweight and obese women, but not normalweight women with PCOS were at an increased risk of T2DM by age 46, suggesting that the risk of T2DM in women with PCOS is mainly due to excess weight, which, as shown in Study I as well as in numerous previous studies (Gambineri et al., 2002; Targher et al., 2009; Yildiz et al., 2008), is a very common disorder in women with PCOS. 63

66 Fig. 15. Prevalence of T2DM and pre-diabetes in normal-weight controls and in women with PCOS at age 46. Fig. 16. Prevalence of T2DM and pre-diabetes in overweight/obese controls and women with PCOS at age 46. Most of the T2DM cases among women with PCOS had already been diagnosed by age 46, as only four nt2dm cases were identified on the basis of the 2h-OGTT results in this study. Of note, the mean BMI of these nt2dm cases was 36.0 ± 5.3 (SD) kg/m 2. Women with PCOS and T2DM also had significantly higher levels of HbA1c compared with women with PCOS and NGT, but HbA1c was 48 mmol/mol (the cut-off value for T2DM) in only four of 23 women with PCOS and T2DM. Our findings suggest that it is not necessary to perform 2h-OGTTs routinely in normal-weight women with PCOS, whereas 2h-OGTT screening should be focused on overweight/obese women with PCOS, and HbA1c seems not to be an accurate tool to detect T2DM in women with PCOS. Our findings are well in line with the recent international evidence-based PCOS guidelines (Teede et al., 2018), 64

67 which recommend the use of 2h-OGTTs in the evaluation of glucose metabolism in overweight/obese women with PCOS. Polycystic ovary syndrome per se was associated with an increased risk of T2DM in overweight/obese women compared with overweight/obese controls (OR 2.45, 95% CI: ). However, we were unable to statistically assess the interaction between PCOS and BMI, due to a lack of normal-weight women with both PCOS and T2DM. In addition, we found that weight gain in early adulthood, between ages 14 and 31, was significantly greater in women with both PCOS and T2DM at age 46 than in women with PCOS and NGT (Figure 17). Fig. 17. Development of BMI in women with PCOS according to glucose metabolism status at age 46. ***p < 0.001, **p < 0.01 and *p < Our findings were recently confirmed in a cross-sectional Nordic multicentre study of 876 women with PCOS from Norway, Sweden, Denmark and Finland who underwent 2h-OGTTs. The results indicated that none of the normal-weight women with PCOS had T2DM, and that elevated BMI was the most important risk factor of abnormal glucose metabolism in PCOS (Pelanis et al., 2017). Likewise, in a nationwide study of Danish women it was found that BMI and fasting blood glucose levels were the best predictors of the development of T2DM in women with PCOS (Rubin, Glintborg, Nybo, Abrahamsen, & Andersen, 2017). In the same study it was also reported that the risk of development of T2DM in normal-weight 65

68 women with PCOS vs. age-matched controls was not elevated (hazard ratio 1.22 [95% CI: ], p = 0.60), whereas in their overweight/obese group, the hazard ratio was as high as 6.57 (95% CI: , p < 0.001) compared with age-matched controls (Glintborg, Rubin, Abrahamsen, & Andersen, 2018). Recently, an updated version (Kakoly et al., 2018) of a previous meta-analysis and systematic review of glucose metabolism in women with PCOS (Moran et al., 2010) was published. Interestingly, this recent update, focusing only on highquality studies, revealed that women with PCOS had a significantly higher prevalence of T2DM compared with non-pcos controls. However, in subgroup analysis there were no significant differences in the prevalence of T2DM among European women with and without PCOS, and the same was found in BMImatched studies (Kakoly et al., 2018). We also found that women with both PCOS and T2DM had comparable levels of serum T at the ages of 31 and 46 as the women with PCOS and NGT, suggesting that hyperandrogenaemia is not associated with abnormal glucose metabolism. This is in line with the results of a previous retrospective cohort study (Mani et al., 2013). In conclusion, our results together with the above-mentioned recent metaanalysis suggest that in Nordic populations, normal-weight women with PCOS do not have an increased prevalence of glucose metabolism abnormalities. However, overweight and obese women with PCOS have a significantly higher risk of T2DM when compared with overweight and obese control women. Weight gain during early adulthood emerged as a significant risk factor of later development of T2DM, emphasizing the important role of weight management, especially during early reproductive years. Our results are best generalised to European populations, as in the aforementioned recent meta-analysis it was reported that ethnicity has a marked impact on the prevalence of abnormal glucose metabolism (Kakoly et al., 2018). 5.3 Hypertension and cardiac structure and function in PCOS (Study III) The main aim of Study III was to investigate the prevalence of hypertension at age 46 in women with PCOS, based on BP levels at clinical examination, hypertension diagnosis set by a physician, and self-reported use of antihypertensive medication. Secondly, we investigated whether the heart s structure and function are altered in women with PCOS at age 46. Our main finding was that women with PCOS were significantly more often hypertensive (and diagnosed as such) and more often used antihypertensive medication compared with controls. Moreover, we found that in 66

69 hypertensive women with PCOS, elevated BP was associated with adverse changes in heart structure and function compared with normotensive women. Hypertension At age 31, women with PCOS (both normal-weight and overweight/obese groups) displayed significantly higher systolic and diastolic BP levels when compared with control women of similar weight (Figure 18). At age 46, normal-weight women with PCOS displayed significantly higher diastolic, but not systolic BP compared with controls, whereas there were no significant differences between overweight/obese PCOS women and overweight/obese controls (Figure 18). Of note, at age 46, the prevalence of self-reported obstructive sleep apnoea was comparable in the PCOS and control groups (1.3% [n = 3/238] vs. 1.2% [n = 19/1543]), indicating that obstructive sleep apnoea did not affect the results. However, the possibility of underdiagnosis of obstructive sleep apnoea should be kept in mind. Our findings are in accordance with the results of two previous (cross-sectional) studies including adolescents (Zachurzok-Buczynska et al., 2011) and young reproductive-aged women (Joham et al., 2015). In these studies it was found that normal-weight women with PCOS had significantly higher BP levels compared with normal-weight controls, whereas overweight/obese PCOS and control groups had similar prevalence rates of hypertension (Joham et al., 2015; Zachurzok- Buczynska et al., 2011). 67

70 Fig. 18. Blood pressure levels in controls and in women with PCOS according to BMI group at ages 31 and 46. **p < 0.01 and *p < Self-reported use of antihypertensive medication was significantly more common among women with PCOS when compared with controls both at ages 31 and 46 (Figure 19), as was the prevalence of self-reported hypertension diagnosis at age 46 (30.1% [n = 72/239] vs. 18.2% [n = 283/1556], p < 0.001). Again, these findings are in line with the results of the Danish nationwide register study, in which it was reported that hypertension diagnosis (based on ICD codes) and use of antihypertensive medication were significantly more common among women with PCOS (mean age 29 years) when compared with the controls (Glintborg et al., 2015). 68

71 Fig. 19. Prevalence of use of antihypertensive medication (%). ***p < and *p < In binary logistic regression analysis, PCOS was significantly associated with hypertension at age 46 independently of overweight/obesity (Figure 20). This difference persisted after adjustment for other potential confounding factors such as smoking, education level, alcohol consumption, physical activity and T2DM. In line with our findings, in a large American register study it was reported that PCOS was significantly associated with hypertension and elevated BP in adjusted regression analysis (Lo et al., 2006). Fig. 20. Binary logistic regression analysis of hypertension (HTA) at age

72 Taken together, women with PCOS were significantly more often hypertensive (and diagnosed as such) and more often used antihypertensive medication compared with controls. Women with PCOS also already had significantly higher BP levels in their 30s when compared with control women in a similar BMI group. Our results, together with those in the aforementioned previous studies, indicate that PCOS is an independent risk factor of hypertension, although weight excess emerged as a major determinant of hypertension in regression analysis. Long-term epidemiological studies have demonstrated a clear relationship between BP and long-term risk of CVD events and mortality in young adults with BP > 130/80 mmhg (Williams et al., 2018). In previous studies it has also been reported that an increase of BP of mmhg has a marked impact on CVD risk at population level (Staessen et al., 1997). Conversely, a reduction of diastolic BP by as little as 2 mmhg can result in a 6% reduction in the risk of coronary heart disease and a 15% reduction in the risk of stroke and transient ischaemic attacks (Cook, Cohen, Hebert, Taylor, & Hennekens, 1995) at a population level. In the same study it was also pointed out that these small changes are also important when BP is at a so-called high-normal level, because a large proportion of the general population displays BP levels in that range, therefore having a marked impact on the attributable risks of stroke and coronary heart disease in general (Cook et al., 1995). In that context, the relatively modest absolute BP difference in our study between PCOS and control women may be of major importance at a population level. An elevated morning BP surge is a predictor of adverse cardiovascular events and, interestingly, in a recent study it was reported that young women with PCOS (mean age 27 years) displayed a significantly greater morning BP surge than control women (Kadi, Avci, Usta, & Demirtaş, 2018). The authors suggested that an exacerbated morning BP surge might be explained by increased IR and increased sympathetic and renin-angiotensin-aldosterone activities in PCOS. Blood pressure and glucose metabolism might also be connected to each other, as vasopressin influences gluconeogenesis and glucogenolysis in the liver as well as insulin release from the pancreas (Bankir, Bichet, & Morgenthaler, 2017). In a study of BMI- and age-matched reproductive-aged controls and women with PCOS it was also reported that copeptin, a surrogate marker of vasopressin, was significantly associated with HOMA-IR, fasting insulin, and free T (Karbek et al., 2014). In conclusion, women with PCOS displayed higher BP than control women, and excess weight seemed to be a major determinant of elevated BP in women with PCOS, although other compex mechanisms may also be implicated. The present 70

73 findings strengthen the need for early systematic screening and efficient treatment of hypertension in women with PCOS. Cardiac structure and function At age 46, women with PCOS displayed small, adverse changes in cardiac structure, such as increases in interventricular septal thickness at diastole (IVSd: 0.91cm ± 0.1 vs cm ± 0.2, p = 0.154), left ventricular mass index (LVMI: ± 32.0 vs ± 38.3, p = 0.283) and left atrial systolic volume (LAESV: ml ± 15.3 vs ml ± 15.1 p = 0.376), compared with the control women. There were no significant differences in systolic (evaluated by left ventricle ejection fraction and global strain) or diastolic function between PCOS and control women. Compared with normotensive control women, hypertensive women with PCOS had significantly higher values of IVSd, LVMI and LAESV (Figure 21). Estimated left ventricular filling pressure (E/e ) was also increased in hypertensive PCOS women, suggesting modest alterations in diastolic function in this PCOS subgroup (Figure 21). Left ventricular systolic function was mildly decreased in hypertensive PCOS women (assessed by Global strain) even though the difference in LVEF was not statistically significant. 71

74 Fig. 21. Echocardiographic parameters in normotensive controls and in hypertensive PCOS women. Increased interventricular septal thickness, left ventricular mass and left atrial systolic volume are typical adaptive changes of the heart in response to elevated BP. These changes are thought to predispose individuals later to heart failure with preserved ejection fraction (Ponikowski et al., 2016). Our findings suggest that elevated BP is the main determinant of adverse cardiac structure in women with PCOS, as normotensive PCOS and control groups showed comparable cardiac structure and function. The relatively small number and young age of the women explain most likely why there were only a few statistically significant differences in the echocardiographic parameters. Our findings are in line with those in three previous studies (Orio et al., 2004; Tíras et al., 1999; Wang et al., 2012), which also revealed changes in left atrial size, left ventricular mass and in the parameters 72

75 describing systolic and diastolic function. However, conflicting results have also been reported (Kosmala et al., 2008; Selcoki et al., 2010), and overall, the roles of confounding factors, such as hypertension, IR and hyperandrogenaemia, are still unclear and need to be investigated in future studies with adequate sample sizes. Moreover, the controversial findings in previous studies might also be explained by differences in age and ethnicity of the study groups as well as by the criteria used for the diagnosis of PCOS. 5.4 PCOS and cardiac autonomic function (Study IV) The main aim of Study IV was to investigate whether premenopausal women with PCOS display impaired HRV as an indicator of impaired cardiac autonomic function, independently of confounding metabolic abnormalities, such as excess weight, abdominal obesity, hyperandrogenaemia, increased BP, dyslipidaemia, and IR. The main finding was that women with PCOS displayed significantly decreased vagal activity, but this seemed to be mainly associated with the metabolic abnormalities linked to the syndrome and not to PCOS itself. At age 46, women with PCOS had significantly higher mean HR values and significantly lower values of RMSSD, LF RRi, HF RRi and BRS, when compared with control women (Figure 22). However, when the analyses were adjusted for BMI, women with PCOS showed significantly smaller values only for RMSSD (p = 0.033) and HF RRi (p = 0.016), when compared with the controls. 73

76 Fig. 22. HRV and baroreflex sensitivity in controls and in women with PCOS in a seated position. HR: Heart rate, RMSSD: The square root of the mean squared differences of successive normal-to-normal RR intervals (RRi), LF RRi : low frequency power, HF RRi : high frequency power, α 1 : short-term fractal-like scaling exponent in detrended fluctuation analysis, SBP: systolic blood pressure, BRS: baroreflex sensitivity. 74

77 The confounding effect of metabolic abnormalities was analysed in multivariate linear regression analysis. The multivariate linear regression model included PCOS, BMI, MAP, FAI, HOMA-IR and triglycerides at age 46 as explanatory variables for RMSSD. The model showed that PCOS, BMI and the FAI were not significantly associated with RMSSD, whereas MAP, HOMA-IR and triglycerides were. Anxiety (assessed by Hopkins Symptom Check List-25, a well-known and widely used symptom inventory) was not included in the final multivariate model, because in the preliminary models, using a step-wise method, anxiety was not significantly associated with RMSSD. Total and calculated free T at age 46 was not significantly associated with RMSSD in linear regression analysis. Serum levels of SHBG were positively associated with RMSSD (B = 0.001, 95% CI: , p < 0.001) and the FAI was negatively associated with it (B = , 95% CI: , p = 0.003), but both associations lost their significance after adjustment for BMI. In line with our findings, in a study of 60 normal-weight women with PCOS (mean age 25 yr) and 60 age-matched normal-weight controls, with comparable HOMA-IR indices, it was reported that there was no difference between the groups in HRV (Kilit & Paşalı Kilit, 2017). These investigators also reported that serum T did not correlate with HRV among the study population, which is also in line with our findings. In another study, including 19 overweight/obese women with PCOS (mean BMI 31 kg/m 2 ) and 19 similar-weight controls it was reported that HRV did not differ between the groups, whereas women with PCOS had elevated MSNA activity (Lambert et al., 2015). In that study, IR and glucose metabolism status seemed to be the main mediators of sympathetic tone, regardless of PCOS status, and T was not associated with MSNA (Lambert et al., 2015). In another crosssectional study, concerning 31 women with PCOS and 30 age-matched controls it was reported that women with PCOS had significantly increased sympathetic activity as well as decreased parasympathetic activity, and they had a significantly more adverse metabolic profile than the control women (Saranya et al., 2014). In a study of 23 young women with PCOS and 25 healthy controls with a similar metabolic profile and normal BMI, comparable HRV parameters between the groups were reported (Özkeçeci et al., 2016). In accordance with our findings, in another study it was found that normalweight PCOS and control groups had comparable HRV parameters, whereas in the overweight/obese group, women with PCOS had significantly impaired HRV compared with those in other groups (Hashim et al., 2015). It is well known that metabolic profiles differ between PCOS phenotypes, suggesting that autonomic 75

78 function might also differ across PCOS phenotypes. Accordingly, women with classic PCOS, i.e. hyperandrogenism and anovulation displayed altered HRV responses to a mental stress test, whereas ovulatory women with PCOS did not significantly differ from controls (Di Domenico et al., 2013). This finding may be due to the fact that women with classic PCOS have a significantly more adverse metabolic and hormonal profile than ovulatory women with PCOS and controls. Last, another study showed that the adverse HRV in women with PCOS was strongly related to BMI, dyslipidaemia, inflammation and IR (de Sá, Joceline Cássia Ferezini et al., 2011). Our findings together with previous reports indicate that women with PCOS display impaired HRV mainly because of the metabolic abnormalities often coexisting with PCOS. In line with ourselves, most investigators have come to the conclusion that obesity, IR, hyperinsulinaemia, and elevated BP are the main determinants of altered autonomic function in women with PCOS. Obesity, and especially excess visceral adipose tissue, correlates with decreased HRV both in the general population (Windham et al., 2012) and in women with PCOS (Hashim et al., 2015). However, obesity seems to increase sympathetic activity in the kidneys and in skeletal muscle vascularity, but reduces it in the heart (Vaz et al., 1997), which might partly explain conflicting results between studies involving HRV and MSNA techniques. There also seems to be a complex relationship between excess adipose tissue, chronic inflammation, IR and sympathetic action in PCOS, but the exact mechanisms among these factors are still unclear (Lansdown & Rees, 2012; Shorakae et al., 2018). Impaired autonomic function has been suggested to be associated with hyperandrogenaemia and elevated BP in PCOS (Perciaccante, Fiorentini, Valente, & Tubani, 2007). The role of hyperandrogenaemia seems to be controversial, as some have found that it was significantly associated with autonomic function (Sverrisdóttir et al., 2008), whereas others (Kilit & Paşalı Kilit, 2017; Lambert et al., 2015; Yildirir et al., 2006), including ourselves, have not. These conflicting results might be due to differences in the assessment and definition of hyperandrogenaemia, different T assays and overall differences in study populations. However, weight loss has been reported to improve HR (and thus vagal activation) and to reduce waist circumference, BP, IR and hyperandrogenaemia in a 10-week prospective clinical intervention including 57 overweight/obese women with PCOS (Thomson et al., 2010). 76

79 5.5 PCOS and cardiovascular disease morbidity (Study III) An important aim of Study III was to investigate whether or not premenopausal women with PCOS have a greater prevalence of CVD events than control women. We found that the prevalence of myocardial infarction was already significantly higher in premenopausal women with PCOS (1.8% vs. 0.5%, respectively, p = 0.034), as was the prevalence of any cardiovascular disease (6.8% vs. 3.4%, p = 0.011) (Figure 23). In addition, by age 46, two women with PCOS, but no woman in the control group, had died as a result of ischaemic heart disease. Fig. 23. Prevalence of acute myocardial infarction and any CVD event in controls and women with PCOS by age 46. Our findings suggest that women with PCOS have an increased prevalence of cardiovascular morbidity. Importantly, we were able to study data on cardiovascular morbidity from high-quality registers that have been shown to have very high coverage and accuracy (88 98%) as regards vascular diseases in Finland (Sund, 2012), and the CVD event diagnoses were also verified by patient chart review. However, these results need to be interpreted with caution, as the number of affected women with overt CVD was very small, and we need to further follow these women to clarify how the prevalence will develop with aging. Our findings are in line with the results of some (Glintborg et al., 2015; Glintborg et al., 2018; Mani et al., 2013) but not all previous studies (Iftikhar et al., 2012; Lo et al., 2006; Merz et al., 2016; Meun et al., 2018; Morgan et al., 2012; Schmidt et al., 2011). These divergent results might be explained by differences in study designs, definitions of PCOS and CVD events, PCOS phenotypes, ages of 77